2019-05-21-modern-data-oriented-programming.html

---
title: "Modern Data Oriented Programming"
venue: "Advances and Challenges in Machine Learning Languages, Centre for Mathematical Sciences, Cambridge"
abstract: "There has been a great deal of interest in probabilistic programs: placing modeling at the heart of programming language. In this talk we set the scene for data oriented programming. Data is a fundamental component of machine learning, yet the availability, quality and discoverability of data are often ignored in formal computer science. While languages for data manipulation exist (for example SQL), they are not suitable for the modern world of machine learning data. Modern data oriented languages should place data at the center of modern digital systems design and provide an infrastructure in which monitoring of data quality and model decision making are automaticaly available. We provide the context for Modern Data Oriented Programming, and give some insight into our initial ideas in this space."
author:
- given: Neil D.
  family: Lawrence
  url: http://inverseprobability.com
  institute: Amazon Cambridge and University of Sheffield
  twitter: lawrennd
  gscholar: r3SJcvoAAAAJ
  orchid: 
date: 2019-05-21
published: 2019-05-21
reveal: 2019-05-21-modern-data-oriented-programming.slides.html
layout: talk
categories:
- notes
---


<div style="display:none">
  $${% include talk-notation.tex %}$$
</div>



    
<div id="modal-frame" class="modal">
  <span class="close" onclick="closeMagnify()">&times;</span>
  <div class="modal-figure">
    <div class="figure-frame">
      <div class="modal-content" id="modal01"></div>
      <!--<img class="modal-content" id="object01">-->
    </div>
    <div class="caption-frame" id="modal-caption"></div>
  </div>
</div>	  

<!-- Front matter -->
<!-- Front matter -->
<!-- Do not edit this file locally. -->
<!-- Do not edit this file locally. -->
<!---->
<!-- Do not edit this file locally. -->
<!--Back matter-->
<!-- Do not edit this file locally. -->
<!-- The last names to be defined. Should be defined entirely in terms of macros from above-->
<!--



## The Gartner Hype Cycle <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/gartner-hype-cycle.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/gartner-hype-cycle.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span>

<div class="figure">
<div class="figure-frame" id="gartner-hype-cycle-figure">
<object class="svgplot " data="../slides/diagrams/Gartner_Hype_Cycle.svg" width="60%" style=" "></object>
</div>
<div class="magnify" id="gartner-hype-cycle-magnify" onclick="magnifyFigure('gartner-hype-cycle')">
<img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex">
</div>
<div class="caption-frame" id="gartner-hype-cycle-caption">
Figure: The Gartner Hype Cycle
</div>
</div>


The [Gartner Hype Cycle](https://en.wikipedia.org/wiki/Hype_cycle) tries to assess where an idea is in terms of maturity and adoption. It splits the evolution of technology into a technological trigger, a peak of expectations followed by a trough of disillusionment and a final ascension into a useful technology. It looks rather like a classical control response to a final set point.














                  
```{.python}
import pods
from ipywidgets import IntSlider
```
```{.python}
pods.notebook.display_plots('ai-bd-dm-dl-ml-google-trends{sample:0>3}.svg', 
                            '../slides/diagrams/data-science/', sample=IntSlider(0, 1, 4, 1))
```






<div class="figure">
<div class="figure-frame" id="ai-bd-dm-dl-ml-gartner-hype-cycle-google-trends-figure">
<object class="svgplot " data="../slides/diagrams/data-science/ai-bd-dm-dl-ml-google-trends.svg" width="80%" style=" "></object>
</div>
<div class="magnify" id="ai-bd-dm-dl-ml-gartner-hype-cycle-google-trends-magnify" onclick="magnifyFigure('ai-bd-dm-dl-ml-gartner-hype-cycle-google-trends')">
<img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex">
</div>
<div class="caption-frame" id="ai-bd-dm-dl-ml-gartner-hype-cycle-google-trends-caption">
Figure: Google trends for 'artificial intelligence', 'big data', 'data mining', 'deep learning', 'machine learning' as different technological terms on the hype cycle.
</div>
</div>


Google trends gives us insight into how far along various technological terms are on the hype cycle.




Examining Google treds for 'artificial intelligence', 'big data', 'data mining', 'deep learning' and 'machine learning' we can see that 'artificial intelligence' *may* be entering a plateau of productivity, 'big data' is entering the trough of disillusionment, and 'data mining' seems to be deeply within the trough. On the other hand 'deep learning' and 'machine learning' appear to be ascending to the peak of inflated expectations having experienced a technology trigger.

For deep learning that technology trigger was the ImageNet result of 2012 [@Krizhevsky:imagenet12]. This step change in performance on object detection in images was achieved through convolutional neural networks, popularly known as 'deep learning'.
-->
<h2 id="the-centrifugal-governor-edit">The Centrifugal Governor <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ai/includes/centrifugal-governor.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ai/includes/centrifugal-governor.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="science-holborn-viaduct-figure" class="figure-frame">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/science-holborn-viaduct.jpg" width="50%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="science-holborn-viaduct-magnify" class="magnify" onclick="magnifyFigure(&#39;science-holborn-viaduct&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="science-holborn-viaduct-caption" class="caption-frame">
<p>Figure: Centrifugal governor as held by “Science” on Holborn Viaduct</p>
</div>
</div>
<h2 id="boulton-and-watts-steam-engine-edit">Boulton and Watt’s Steam Engine <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ai/includes/watt-steam-engine.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ai/includes/watt-steam-engine.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="steam-engine-boulton-watt-figure" class="figure-frame">
<div class="centered" style="">
<img class="negate" src="../slides/diagrams/SteamEngine_Boulton&Watt_1784.png" width="70%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="steam-engine-boulton-watt-magnify" class="magnify" onclick="magnifyFigure(&#39;steam-engine-boulton-watt&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="steam-engine-boulton-watt-caption" class="caption-frame">
<p>Figure: Watt’s Steam Engine which made Steam Power Efficient and Practical.</p>
</div>
</div>
<p>James Watt’s steam engine contained an early machine learning device. In the same way that modern systems are component based, his engine was composed of components. One of which is a speed regulator sometimes known as <em>Watt’s governor</em>. The two balls in the center of the image, when spun fast, rise, and through a linkage mechanism.</p>
<p>The centrifugal governor was made famous by Boulton and Watt when it was deployed in the steam engine. Studying stability in the governor is the main subject of James Clerk Maxwell’s paper on the theoretical analysis of governors <span class="citation" data-cites="Maxwell:governors1867">(Maxwell 1867)</span>. This paper is a founding paper of control theory. In an acknowledgment of its influence, Wiener used the name <a href="https://en.wikipedia.org/wiki/Cybernetics"><em>cybernetics</em></a> to describe the field of control and communication in animals and the machine <span class="citation" data-cites="Wiener:cybernetics48">(Wiener 1948)</span>. Cybernetics is the Greek word for governor, which comes from the latin for helmsman.</p>
<p>A governor is one of the simplest artificial intelligence systems. It senses the speed of an engine, and acts to change the position of the valve on the engine to slow it down.</p>
<p>Although it’s a mechanical system a governor can be seen as automating a role that a human would have traditionally played. It is an early example of artificial intelligence.</p>
<p>The centrifugal governor has several parameters, the weight of the balls used, the length of the linkages and the limits on the balls movement.</p>
<p>Two principle differences exist between the centrifugal governor and artificial intelligence systems of today.</p>
<ol type="1">
<li>The centrifugal governor is a physical system and it is an integral part of a wider physical system that it regulates (the engine).</li>
<li>The parameters of the governor were set by hand, our modern artificial intelligence systems have their parameters set by <em>data</em>.</li>
</ol>
<div class="figure">
<div id="centrifugal-governor-figure" class="figure-frame">
<div class="centered" style="">
<img class="negate" src="../slides/diagrams/Centrifugal_governor.png" width="70%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="centrifugal-governor-magnify" class="magnify" onclick="magnifyFigure(&#39;centrifugal-governor&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="centrifugal-governor-caption" class="caption-frame">
<p>Figure: The centrifugal governor, an early example of a decision making system. The parameters of the governor include the lengths of the linkages (which effect how far the throttle opens in response to movement in the balls), the weight of the balls (which effects inertia) and the limits of to which the balls can rise.</p>
</div>
</div>
<p>This has the basic components of sense and act that we expect in an intelligent system, and this system saved the need for a human operator to manually adjust the system in the case of overspeed. Overspeed has the potential to destroy an engine, so the governor operates as a safety device.</p>
<p>The first wave of automation did bring about sabotoage as a worker’s response. But if machinery was sabotaged, for example, if the linkage between sensor (the spinning balls) and action (the valve closure) was broken, this would be obvious to the engine operator at start up time. The machine could be repaired before operation.</p>
<h2 id="what-is-machine-learning-edit">What is Machine Learning? <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ml/includes/what-is-ml-2.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ml/includes/what-is-ml-2.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>Machine learning allows us to extract knowledge from data to form a prediction.</p>
<p><br /><span class="math display">$$\text{data} + \text{model} \xrightarrow{\text{compute}} \text{prediction}$$</span><br /></p>
<p>A machine learning prediction is made by combining a model with data to form the prediction. The manner in which this is done gives us the machine learning <em>algorithm</em>.</p>
<p>Machine learning models are <em>mathematical models</em> which make weak assumptions about data, e.g. smoothness assumptions. By combining these assumptions with the data, we observe we can interpolate between data points or, occasionally, extrapolate into the future.</p>
<p>Machine learning is a technology which strongly overlaps with the methodology of statistics. From a historical/philosophical view point, machine learning differs from statistics in that the focus in the machine learning community has been primarily on accuracy of prediction, whereas the focus in statistics is typically on the interpretability of a model and/or validating a hypothesis through data collection.</p>
<p>The rapid increase in the availability of compute and data has led to the increased prominence of machine learning. This prominence is surfacing in two different but overlapping domains: data science and artificial intelligence.</p>
<h2 id="from-model-to-decision-edit">From Model to Decision <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ml/includes/what-is-ml-end-to-end.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ml/includes/what-is-ml-end-to-end.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>The real challenge, however, is end-to-end decision making. Taking information from the environment and using it to drive decision making to achieve goals.</p>
<h2 id="artificial-intelligence-and-data-science-edit">Artificial Intelligence and Data Science <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ai/includes/ai-vs-data-science-2.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ai/includes/ai-vs-data-science-2.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>Artificial intelligence has the objective of endowing computers with human-like intelligent capabilities. For example, understanding an image (computer vision) or the contents of some speech (speech recognition), the meaning of a sentence (natural language processing) or the translation of a sentence (machine translation).</p>
<h3 id="supervised-learning-for-ai">Supervised Learning for AI</h3>
<p>The machine learning approach to artificial intelligence is to collect and annotate a large data set from humans. The problem is characterized by input data (e.g. a particular image) and a label (e.g. is there a car in the image yes/no). The machine learning algorithm fits a mathematical function (I call this the <em>prediction function</em>) to map from the input image to the label. The parameters of the prediction function are set by minimizing an error between the function’s predictions and the true data. This mathematical function that encapsulates this error is known as the <em>objective function</em>.</p>
<p>This approach to machine learning is known as <em>supervised learning</em>. Various approaches to supervised learning use different prediction functions, objective functions or different optimization algorithms to fit them.</p>
<p>For example, <em>deep learning</em> makes use of <em>neural networks</em> to form the predictions. A neural network is a particular type of mathematical function that allows the algorithm designer to introduce invariances into the function.</p>
<p>An invariance is an important way of including prior understanding in a machine learning model. For example, in an image, a car is still a car regardless of whether it’s in the upper left or lower right corner of the image. This is known as translation invariance. A neural network encodes translation invariance in <em>convolutional layers</em>. Convolutional neural networks are widely used in image recognition tasks.</p>
<p>An alternative structure is known as a recurrent neural network (RNN). RNNs neural networks encode temporal structure. They use auto regressive connections in their hidden layers, they can be seen as time series models which have non-linear auto-regressive basis functions. They are widely used in speech recognition and machine translation.</p>
<p>Machine learning has been deployed in Speech Recognition (e.g. Alexa, deep neural networks, convolutional neural networks for speech recognition), in computer vision (e.g. Amazon Go, convolutional neural networks for person recognition and pose detection).</p>
<p>The field of data science is related to AI, but philosophically different. It arises because we are increasingly creating large amounts of data through <em>happenstance</em> rather than active collection. In the modern era data is laid down by almost all our activities. The objective of data science is to extract insights from this data.</p>
<p>Classically, in the field of statistics, data analysis proceeds by assuming that the question (or scientific hypothesis) comes before the data is created. E.g., if I want to determine the effectiveness of a particular drug, I perform a <em>design</em> for my data collection. I use foundational approaches such as randomization to account for confounders. This made a lot of sense in an era where data had to be actively collected. The reduction in cost of data collection and storage now means that many data sets are available which weren’t collected with a particular question in mind. This is a challenge because bias in the way data was acquired can corrupt the insights we derive. We can perform randomized control trials (or A/B tests) to verify our conclusions, but the opportunity is to use data science techniques to better guide our question selection or even answer a question without the expense of a full randomized control trial (referred to as A/B testing in modern internet parlance).</p>
<h2 id="amazon-bits-and-atoms">Amazon: Bits and Atoms</h2>
<!--
include{../_ai/includes/embodiment-factors.md}
include{_data-science/includes/evolved-relationship.md}
include{_ml/includes/what-does-machine-learning-do.md}

newslide{Deep Learning}

* These are interpretable models: vital for disease etc.

* Modern machine learning methods are less interpretable

* Example: face recognition

include{_ml/includes/deep-learning-overview.md}-->
<!--include{_gp/includes/gp-intro-very-short.md}-->
<!--include{_deepgp/includes/deep-olympic.md}-->
<!--
include{_data-science/includes/a-time-for-professionalisation.md}
include{_data-science/includes/the-data-crisis.md} 

newslide{Rest of this Talk: Two Areas of Focus}

* Reusability of Data
* Deployment of Machine Learning Systems

newslide{Rest of this Talk: Two Areas of Focus}

* <s>Reusability of Data</s>
* Deployment of Machine Learning Systems

include{_data-science/includes/data-readiness-levels.md}



* Challenges in deploying AI.
* Currently this is in the form of "machine learning systems"



* Fog computing: barrier between cloud and device blurring.
    * Computing on the Edge
* Complex feedback between algorithm and implementation
  


* Major new challenge for systems designers.
* Internet of Intelligence but currently:
    * AI systems are *fragile*





-->
<h2 id="supply-chain-edit">Supply Chain <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_supply-chain/includes/supply-chain.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_supply-chain/includes/supply-chain.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="packhorse-bridge-burbage-brook-figure" class="figure-frame">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/supply-chain/packhorse-bridge-burbage-brook.jpg" width="80%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="packhorse-bridge-burbage-brook-magnify" class="magnify" onclick="magnifyFigure(&#39;packhorse-bridge-burbage-brook&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="packhorse-bridge-burbage-brook-caption" class="caption-frame">
<p>Figure: Packhorse Bridge under Burbage Edge. This packhorse route climbs steeply out of Hathersage and heads towards Sheffield. Packhorses were the main route for transporting goods across the Peak District. The high cost of transport is one driver of the ‘smith’ model, where there is a local skilled person responsible for assembling or creating goods (e.g. a blacksmith).</p>
</div>
</div>
<p>On Sunday mornings in Sheffield, I often used to run across Packhorse Bridge in Burbage valley. The bridge is part of an ancient network of trails crossing the Pennines that, before Turnpike roads arrived in the 18th century, was the main way in which goods were moved. Given that the moors around Sheffield were home to sand quarries, tin mines, lead mines and the villages in the Derwent valley were known for nail and pin manufacture, this wasn’t simply movement of agricultural goods, but it was the infrastructure for industrial transport.</p>
<p>The profession of leading the horses was known as a Jagger and leading out of the village of Hathersage is Jagger’s Lane, a trail that headed underneath Stanage Edge and into Sheffield.</p>
<p>The movement of goods from regions of supply to areas of demand is fundamental to our society. The physical infrastructure of supply chain has evolved a great deal over the last 300 years.</p>
<h2 id="cromford-edit">Cromford <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_supply-chain/includes/cromford.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_supply-chain/includes/cromford.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="cromford-mill-figure" class="figure-frame">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/supply-chain/cromford-mill.jpg" width="80%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="cromford-mill-magnify" class="magnify" onclick="magnifyFigure(&#39;cromford-mill&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="cromford-mill-caption" class="caption-frame">
<p>Figure: Richard Arkwright is regarded of the founder of the modern factory system. Factories exploit distribution networks to centralize production of goods. Arkwright located his factory in Cromford due to proximity to Nottingham Weavers (his market) and availability of water power from the tributaries of the Derwent river. When he first arrived there was almost no transportation network. Over the following 200 years The Cromford Canal (1790s), a Turnpike (now the A6, 1816-18) and the High Peak Railway (now closed, 1820s) were all constructed to improve transportation access as the factory blossomed.</p>
</div>
</div>
<p>Richard Arkwright is known as the father of the modern factory system. In 1771 he set up a <a href="https://en.wikipedia.org/wiki/Cromford_Mill">Mill</a> for spinning cotton yarn in the village of Cromford, in the Derwent Valley. The Derwent valley is relatively inaccessible. Raw cotton arrived in Liverpool from the US and India. It needed to be transported on packhorse across the bridleways of the Pennines. But Cromford was a good location due to proximity to Nottingham, where weavers where consuming the finished thread, and the availability of water power from small tributaries of the Derwent river for Arkwright’s <a href="https://en.wikipedia.org/wiki/Spinning_jenny">water frames</a> which automated the production of yarn from raw cotton.</p>
<p>By 1794 the <a href="https://en.wikipedia.org/wiki/Cromford_Canal">Cromford Canal</a> was opened to bring coal in to Cromford and give better transport to Nottingham. The construction of the canals was driven by the need to improve the transport infrastructure, facilitating the movement of goods across the UK. Canals, roads and railways were initially constructed by the economic need for moving goods. To improve supply chain.</p>
<p>The A6 now does pass through Cromford, but at the time he moved there there was merely a track. The High Peak Railway was opened in 1832, it is now converted to the High Peak Trail, but it remains the highest railway built in Britain.</p>
<p><span class="citation" data-cites="Cooper:transformation91">Cooper (1991)</span></p>
<h2 id="containerization-edit">Containerization <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_supply-chain/includes/containerisation.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_supply-chain/includes/containerisation.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="container-2539942_1920-figure" class="figure-frame">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/supply-chain/container-2539942_1920.jpg" width="80%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="container-2539942_1920-magnify" class="magnify" onclick="magnifyFigure(&#39;container-2539942_1920&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="container-2539942_1920-caption" class="caption-frame">
<p>Figure: The container is one of the major drivers of globalization, and arguably the largest agent of social change in the last 100 years. It reduces the cost of transportation, significantly changing the appropriate topology of distribution networks. The container makes it possible to ship goods halfway around the world for cheaper than it costs to process those goods, leading to an extended distribution topology.</p>
</div>
</div>
<p>Containerization has had a dramatic effect on global economics, placing many people in the developing world at the end of the supply chain.</p>
<div class="figure">
<div id="wild-alaskan-cod-figure" class="figure-frame">
<table>
<tr>
<td width="45%">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/supply-chain/wild-alaskan-cod.jpg" width="90%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</td>
<td width="45%">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/supply-chain/wild-alaskan-cod-made-in-china.jpg" width="90%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</td>
</tr>
</table>
</div>
<div id="wild-alaskan-cod-magnify" class="magnify" onclick="magnifyFigure(&#39;wild-alaskan-cod&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="wild-alaskan-cod-caption" class="caption-frame">
<p>Figure: Wild Alaskan Cod, being solid in the Pacific Northwest, that is a product of China. It is cheaper to ship the deep frozen fish thousands of kilometers for processing than to process locally.</p>
</div>
</div>
<p>For example, you can buy Wild Alaskan Cod fished from Alaska, processed in China, sold in North America. This is driven by the low cost of transport for frozen cod vs the higher relative cost of cod processing in the US versus China. Similarly, <a href="https://www.telegraph.co.uk/news/uknews/1534286/12000-mile-trip-to-have-seafood-shelled.html" target="_blank" >Scottish prawns are also processed in China for sale in the UK.</a></p>
<p>This effect on cost of transport vs cost of processing is the main driver of the topology of the modern supply chain and the associated effect of globalization. If transport is much cheaper than processing, then processing will tend to agglomerate in places where processing costs can be minimized.</p>
<p>Large scale global economic change has principally been driven by changes in the technology that drives supply chain.</p>
<p>Supply chain is a large-scale automated decision making network. Our aim is to make decisions not only based on our models of customer behavior (as observed through data), but also by accounting for the structure of our fulfilment center, and delivery network.</p>
<p>Many of the most important questions in supply chain take the form of counterfactuals. E.g. “What would happen if we opened a manufacturing facility in Cambridge?” A counter factual is a question that implies a mechanistic understanding of a system. It goes beyond simple smoothness assumptions or translation invariants. It requires a physical, or <em>mechanistic</em> understanding of the supply chain network. For this reason, the type of models we deploy in supply chain often involve simulations or more mechanistic understanding of the network.</p>
<p>In supply chain Machine Learning alone is not enough, we need to bridge between models that contain real mechanisms and models that are entirely data driven.</p>
<p>This is challenging, because as we introduce more mechanism to the models we use, it becomes harder to develop efficient algorithms to match those models to data.</p>
<p>So many examples in terms of the need for intelligent decision making are based around the challenge of moving goods/energy/compute/water/medicines/drivers/people from where it is to where it needs to be. In other words matching supply with demand. That led me to a motto I developed while working in Amazon’s supply chain.</p>
<blockquote>
<p>Solve Supply Chain, then solve everything else.</p>
</blockquote>
<!--include{_ml/includes/or-control-econometrics-statistics-ml.md}-->
<h2 id="the-three-ds-of-machine-learning-systems-design-edit">The Three Ds of Machine Learning Systems Design <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/talks/edit/gh-pages/modern-data-oriented-programming.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/talks/edit/gh-pages/modern-data-oriented-programming.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="experiment-analyze-design-figure" class="figure-frame">
<object class="svgplot " data="../slides/diagrams/ml/experiment-analyze-design.svg" width="50%" style=" ">
</object>
</div>
<div id="experiment-analyze-design-magnify" class="magnify" onclick="magnifyFigure(&#39;experiment-analyze-design&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="experiment-analyze-design-caption" class="caption-frame">
<p>Figure: The experiment, analyze, design flywheel for scientific innovation.</p>
</div>
</div>
<blockquote>
<p>We don’t know what science we’ll want to do in 5 years time, but we won’t want slower experiments, we won’t want more expensive experiments and we won’t want a narrower selection of experiments.</p>
</blockquote>
<ul>
<li>Faster, cheaper and more diverse experiments.</li>
<li>Better ecosystems for experimentation.</li>
<li>Data oriented architectures.</li>
</ul>
<p>We can characterize the challenges for integrating machine learning within our systems as the three Ds. Decomposition, Data and Deployment.</p>
<p>The first two components <em>decomposition</em> and <em>data</em> are interlinked, but we will first outline the decomposition challenge. Below we will mainly focus on <em>supervised learning</em> because this is arguably the technology that is best understood within machine learning.</p>
<h2 id="data-edit">Data <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ml/includes/ml-data-challenge.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ml/includes/ml-data-challenge.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>It is difficult to overstate the importance of data. It is half of the equation for machine learning but is often utterly neglected. We can speculate that there are two reasons for this. Firstly, data cleaning is perceived as tedious. It doesn’t seem to consist of the same intellectual challenges that are inherent in constructing complex mathematical models and implementing them in code. Secondly, data cleaning is highly complex, it requires a deep understanding of how machine learning systems operate and good intuitions about the data itself, the domain from which data is drawn (e.g. Supply Chain) and what downstream problems might be caused by poor data quality.</p>
<p>A consequence of these two reasons, data cleaning seems difficult to formulate into a readily teachable set of principles. As a result, it is heavily neglected in courses on machine learning and data science. Despite data being half the equation, most University courses spend little to no time on its challenges.</p>
<p>Anecdotally, talking to data modelling scientists. Most say they spend 80% of their time acquiring and cleaning data. This is precipitating what I refer to as the “data crisis”. This is an analogy with software. The “software crisis” was the phenomenon of inability to deliver software solutions due to increasing complexity of implementation. There was no single shot solution for the software crisis, it involved better practice (scrum, test orientated development, sprints, code review), improved programming paradigms (object orientated, functional) and better tools (CVS, then SVN, then git).</p>
<h2 id="the-data-crisis-edit">The Data Crisis <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/the-data-crisis.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/the-data-crisis.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>Anecdotally, talking to data modelling scientists. Most say they spend 80% of their time acquiring and cleaning data. This is precipitating what I refer to as the “data crisis”. This is an analogy with software. The “software crisis” was the phenomenon of inability to deliver software solutions due to increasing complexity of implementation. There was no single shot solution for the software crisis, it involved better practice (scrum, test orientated development, sprints, code review), improved programming paradigms (object orientated, functional) and better tools (CVS, then SVN, then git).</p>
<p>However, these challenges aren’t new, they are merely taking a different form. From the computer’s perspective software <em>is</em> data. The first wave of the data crisis was known as the <em>software crisis</em>.</p>
<h3 id="the-software-crisis">The Software Crisis</h3>
<p>In the late sixties early software programmers made note of the increasing costs of software development and termed the challenges associated with it as the “<a href="https://en.wikipedia.org/wiki/Software_crisis">Software Crisis</a>”. Edsger Dijkstra referred to the crisis in his 1972 Turing Award winner’s address.</p>
<blockquote>
<p>The major cause of the software crisis is that the machines have become several orders of magnitude more powerful! To put it quite bluntly: as long as there were no machines, programming was no problem at all; when we had a few weak computers, programming became a mild problem, and now we have gigantic computers, programming has become an equally gigantic problem.</p>
<p>Edsger Dijkstra (1930-2002), The Humble Programmer</p>
</blockquote>
<blockquote>
<p>The major cause of the data crisis is that machines have become more interconnected than ever before. Data access is therefore cheap, but data quality is often poor. What we need is cheap high-quality data. That implies that we develop processes for improving and verifying data quality that are efficient.</p>
<p>There would seem to be two ways for improving efficiency. Firstly, we should not duplicate work. Secondly, where possible we should automate work.</p>
</blockquote>
<p>What I term “The Data Crisis” is the modern equivalent of this problem. The quantity of modern data, and the lack of attention paid to data as it is initially “laid down” and the costs of data cleaning are bringing about a crisis in data-driven decision making. This crisis is at the core of the challenge of <em>technical debt</em> in machine learning <span class="citation" data-cites="Sculley:debt15">(Sculley et al. 2015)</span>.</p>
<p>Just as with software, the crisis is most correctly addressed by ‘scaling’ the manner in which we process our data. Duplication of work occurs because the value of data cleaning is not correctly recognised in management decision making processes. Automation of work is increasingly possible through techniques in “artificial intelligence”, but this will also require better management of the data science pipeline so that data about data science (meta-data science) can be correctly assimilated and processed. The Alan Turing institute has a program focussed on this area, <a href="https://www.turing.ac.uk/research_projects/artificial-intelligence-data-analytics/">AI for Data Analytics</a>.</p>
<p>Data is the new software, and the data crisis is already upon us. It is driven by the cost of cleaning data, the paucity of tools for monitoring and maintaining our deployments, the provenance of our models (e.g. with respect to the data they’re trained on).</p>
<p>Three principal changes need to occur in response. They are cultural and infrastructural.</p>
<h3 id="the-data-first-paradigm">The Data First Paradigm</h3>
<p>First of all, to excel in data driven decision making we need to move from a <em>software first</em> paradigm to a <em>data first</em> paradigm. That means refocusing on data as the product. Software is the intermediary to producing the data, and its quality standards must be maintained, but not at the expense of the data we are producing. Data cleaning and maintenance need to be prized as highly as software debugging and maintenance. Instead of <em>software</em> as a service, we should refocus around <em>data</em> as a service. This first change is a cultural change in which our teams think about their outputs in terms of data. Instead of decomposing our systems around the software components, we need to decompose them around the data generating and consuming components.<a href="#fn1" class="footnote-ref" id="fnref1"><sup>1</sup></a> Software first is only an intermediate step on the way to becoming <em>data first</em>. It is a necessary, but not a sufficient condition for efficient machine learning systems design and deployment. We must move from <em>software orientated architecture</em> to a <em>data orientated architecture</em>.</p>
<h3 id="data-quality">Data Quality</h3>
<p>Secondly, we need to improve our language around data quality. We cannot assess the costs of improving data quality unless we generate a language around what data quality means. <!--Data Readiness Levels[^data-readiness-levels] are an assessment of data quality that is based on the usage to which data is
put.

[^data-readiness-levels]: [Data Readiness Levels](http://inverseprobability.com/2017/01/12/data-readiness-levels) [@Lawrence:drl17] are an attempt to develop a language around data quality that can bridge the gap between technical solutions and decision makers such as managers and project planners. They are inspired by Technology Readiness Levels which attempt to quantify the readiness of technologies for deployment.--></p>
<h3 id="data-readiness-levels-edit">Data Readiness Levels <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-readiness-levels-short.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-readiness-levels-short.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h3>
<p><a href="http://inverseprobability.com/2017/01/12/data-readiness-levels">Data Readiness Levels</a> <span class="citation" data-cites="Lawrence:drl17">(Lawrence 2017)</span> are an attempt to develop a language around data quality that can bridge the gap between technical solutions and decision makers such as managers and project planners. The are inspired by Technology Readiness Levels which attempt to quantify the readiness of technologies for deployment.</p>
<p>See this blog onblog post on <a href="http://inverseprobability.com/2017/01/12/data-readiness-levels">Data Readiness Levels</a>..</p>
<h3 id="three-grades-of-data-readiness-edit">Three Grades of Data Readiness <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/three-grades-of-data-readiness.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/three-grades-of-data-readiness.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h3>
<p>Data-readiness describes, at its coarsest level, three separate stages of data graduation.</p>
<ul>
<li>Grade C - accessibility
<ul>
<li>Transition: data becomes electronically available</li>
</ul></li>
<li>Grade B - validity
<ul>
<li>Transition: pose a question to the data.</li>
</ul></li>
<li>Grade A - usability</li>
</ul>
<p>The important definitions are at the transition. The move from Grade C data to Grade B data is delimited by the <em>electronic availability</em> of the data. The move from Grade B to Grade A data is delimited by posing a question or task to the data <span class="citation" data-cites="Lawrence:drl17">(Lawrence 2017)</span>.</p>
<p><strong>Recommendation</strong>: Build a shared understanding of the language of data readiness levels for use in planning documents and costing of data cleaning and the benefits of reusing cleaned data.</p>
<h3 id="move-beyond-software-engineering-to-data-engineering">Move Beyond Software Engineering to Data Engineering</h3>
<p>Thirdly, we need to improve our mental model of the separation of data science from applied science. A common trap in our thinking around data is to see data science (and data engineering, data preparation) as a sub-set of the software engineer’s or applied scientist’s skill set. As a result, we recruit and deploy the wrong type of resource. Data preparation and question formulation is superficially similar to both because of the need for programming skills, but the day to day problems faced are very different.</p>
<h2 id="combining-data-and-systems-design-edit">Combining Data and Systems Design <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ml/includes/ml-combining-data-and-systems-design-challenge.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ml/includes/ml-combining-data-and-systems-design-challenge.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<h2 id="data-science-as-debugging-edit">Data Science as Debugging <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-science-as-debugging.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-science-as-debugging.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>One challenge for existing information technology professionals is realizing the extent to which a software ecosystem based on data differs from a classical ecosystem. In particular, by ingesting data we bring unknowns/uncontrollables into our decision-making system. This presents opportunity for adversarial exploitation and unforeseen operation.</p>
<p>blog post on <a href="http://inverseprobability.com/2017/03/14/data-science-as-debugging">Data Science as Debugging</a>.</p>
<p>Starting with the analysis of a data set, the nature of data science is somewhat difference from classical software engineering.</p>
<p>One analogy I find helpful for understanding the depth of change we need is the following. Imagine as a software engineer, you find a USB stick on the ground. And for some reason you <em>know</em> that on that USB stick is a particular API call that will enable you to make a significant positive difference on a business problem. You don’t know which of the many library functions on the USB stick are the ones that will help. And it could be that some of those library functions will hinder, perhaps because they are just inappropriate or perhaps because they have been placed there maliciously. The most secure thing to do would be to <em>not</em> introduce this code into your production system at all. But what if your manager told you to do so, how would you go about incorporating this code base?</p>
<p>The answer is <em>very</em> carefully. You would have to engage in a process more akin to debugging than regular software engineering. As you understood the code base, for your work to be reproducible, you should be documenting it, not just what you discovered, but how you discovered it. In the end, you typically find a single API call that is the one that most benefits your system. But more thought has been placed into this line of code than any line of code you have written before.</p>
<p>An enormous amount of debugging would be required. As the nature of the code base is understood, software tests to verify it also need to be constructed. At the end of all your work, the lines of software you write to actually interact with the software on the USB stick are likely to be minimal. But more thought would be put into those lines than perhaps any other lines of code in the system.</p>
<p>Even then, when your API code is introduced into your production system, it needs to be deployed in an environment that monitors it. We cannot rely on an individual’s decision making to ensure the quality of all our systems. We need to create an environment that includes quality controls, checks and bounds, tests, all designed to ensure that assumptions made about this foreign code base are remaining valid.</p>
<p>This situation is akin to what we are doing when we incorporate data in our production systems. When we are consuming data from others, we cannot assume that it has been produced in alignment with our goals for our own systems. Worst case, it may have been adversarially produced. A further challenge is that data is dynamic. So, in effect, the code on the USB stick is evolving over time.</p>
<p>It might see that this process is easy to formalize now, we simply need to check what the formal software engineering process is for debugging, because that is the current software engineering activity that data science is closest to. But when we look for a formalization of debugging, we find that there is none. Indeed, modern software engineering mainly focusses on ensuring that code is written without bugs in the first place.</p>
<p><strong>Recommendation</strong>: Anecdotally, resolving a machine learning challenge requires 80% of the resource to be focused on the data and perhaps 20% to be focused on the model. But many companies are too keen to employ machine learning engineers who focus on the models, not the data. We should change our hiring priorities and training. Universities cannot provide the understanding of how to data-wrangle. Companies must fill this gap.</p>
<div class="figure">
<div id="derwent-valley-resevoir-figure" class="figure-frame">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/data-science/water-bridge-hill-transport-arch-calm-544448-pxhere.jpg" width="80%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="derwent-valley-resevoir-magnify" class="magnify" onclick="magnifyFigure(&#39;derwent-valley-resevoir&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="derwent-valley-resevoir-caption" class="caption-frame">
<p>Figure: A reservoir of data has more value if the data is consumable. The data crisis can only be addressed if we focus on outputs rather than inputs.</p>
</div>
</div>
<div class="figure">
<div id="lake-district-stream-figure" class="figure-frame">
<div class="centered centered" style="">
<img class="" src="../slides/diagrams/data-science/1024px-Lake_District_picture.jpg" width="80%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="lake-district-stream-magnify" class="magnify" onclick="magnifyFigure(&#39;lake-district-stream&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="lake-district-stream-caption" class="caption-frame">
<p>Figure: For a data first architecture we need to clean our data at source, rather than individually cleaning data for each task. This involves a shift of focus from our inputs to our outputs. We should provide data streams that are consumable by many teams without purification.</p>
</div>
</div>
<p><strong>Recommendation</strong>: We need to share best practice around data deployment across our teams. We should make best use of our processes where applicable, but we need to develop them to become <em>data first</em> organizations. Data needs to be cleaned at <em>output</em> not at <em>input</em>.</p>
<h2 id="outlook-for-machine-learning">Outlook for Machine Learning</h2>
<p>Machine learning has risen to prominence as an approach to <em>scaling</em> our activities. For us to continue to automate in the manner we have over the last two decades, we need to make more use of computer-based automation. Machine learning is allowing us to automate processes that were out of reach before.</p>
<h2 id="data-oriented-architectures-edit">Data Oriented Architectures <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-oriented-architectures.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-oriented-architectures.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>In a streaming architecture we shift from management of services, to management of data streams. Instead of worrying about availability of the services we shift to worrying about the quality of the data those services are producing.</p>
<p>Historically we’ve been <em>software first</em>, this is a necessary but insufficient condition for <em>data first</em>. We need to move from software-as-a-service to data-as-a-service, from service oriented architectures to <em>data oriented architectures</em>.</p>
<h2 id="streaming-system">Streaming System</h2>
<p>Characteristics of a streaming system include a move from <em>pull</em> updates to <em>push</em> updates, i.e. the computation is driven by a change in the input data rather than the service calling for input data when it decides to run a computation. Streaming systems operate on ‘rows’ of the data rather than ‘columns’. This is because the full column isn’t normally available as it changes over time. As an important design principle, the services themselves are stateless, they take their state from the streaming ecosystem. This ensures the inputs and outputs of given computations are easy to declare. As a result, persistence of the data is also handled by the streaming ecosystem and decisions around data retention or recomputation can be taken at the systems level rather than the component level.</p>
<p><strong>Recommendation</strong>: We should consider a major re-architecting of systems around our services. In particular we should scope the use of a <em>streaming architecture</em> (such as Apache Kafka) that ensures data persistence and enables asynchronous operation of our systems.<a href="#fn2" class="footnote-ref" id="fnref2"><sup>2</sup></a> This would enable the provision of QC streams, and real time dash boards as well as hypervisors.</p>
<p>Importantly a streaming architecture implies the services we build are <em>stateless</em>, internal state is deployed on streams alongside external state. This allows for rapid assessment of other services’ data.</p>
<h2 id="apache-flink-edit">Apache Flink <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/apache-flink.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/apache-flink.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p><a href="https://en.wikipedia.org/wiki/Apache_Flink">Apache Flink</a> is a stream processing framework. Flink is a foundation for event driven processing. This gives a high throughput and low latency framework that operates on dataflows.</p>
<p>Data storage is handled by other systems such as Apache Kafka or AWS Kinesis.</p>
<pre><code>stream.join(otherStream)
    .where(&lt;KeySelector&gt;)
    .equalTo(&lt;KeySelector&gt;)
    .window(&lt;WindowAssigner&gt;)
    .apply(&lt;JoinFunction&gt;)</code></pre>
<p>Apache Flink allows operations on streams. For example, the join operation above. In a traditional data base management system, this join operation may be written in SQL and called on demand. In a streaming ecosystem, computations occur as and when the streams update.</p>
<p>The join is handled by the ecosystem surrounding the business logic.</p>
<h2 id="milan-edit">Milan <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/milan.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/milan.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>Milan is a data-oriented programming language and runtime infrastructure.</p>
<p><a href="https://github.com/amzn/milan" class="uri">https://github.com/amzn/milan</a></p>
<p>The Milan language is a DSL embedded in Scala. The output is an intermediate language that can be compiled to run on different target platforms. Currently there exists a single compiler that produces Flink applications.</p>
<p>The Milan runtime infrastructure compiles and runs Milan applications on a Flink cluster.</p>
<h2 id="trading-system">Trading System</h2>
<p>As a simple example we’ll consider a high frequency trading system. Anne wishes to build a share trading system. She has access to a high frequency trading system which provides prices and allows trades at millisecond intervals. She wishes to build an automated trading system.</p>
<p>Let’s assume that price trading data is available as a data stream. But the price now is not the only information that Anne needs, she needs an estimate of the price in the future.</p>
<div class="sourceCode" id="cb2"><pre class="sourceCode python"><code class="sourceCode python"><a class="sourceLine" id="cb2-1" data-line-number="1"><span class="im">import</span> pandas <span class="im">as</span> pd</a>
<a class="sourceLine" id="cb2-2" data-line-number="2"><span class="im">import</span> numpy <span class="im">as</span> np</a>
<a class="sourceLine" id="cb2-3" data-line-number="3"><span class="im">import</span> os</a></code></pre></div>
<div class="sourceCode" id="cb3"><pre class="sourceCode python"><code class="sourceCode python"><a class="sourceLine" id="cb3-1" data-line-number="1"><span class="co"># Generate an artificial trading stream</span></a>
<a class="sourceLine" id="cb3-2" data-line-number="2">days<span class="op">=</span>pd.date_range(start<span class="op">=</span><span class="st">&#39;21/5/2017&#39;</span>, end<span class="op">=</span><span class="st">&#39;21/05/2020&#39;</span>)</a>
<a class="sourceLine" id="cb3-3" data-line-number="3">z <span class="op">=</span> np.random.randn(<span class="bu">len</span>(days), <span class="dv">1</span>)</a>
<a class="sourceLine" id="cb3-4" data-line-number="4">x <span class="op">=</span> z.cumsum()<span class="op">+</span><span class="dv">400</span></a></code></pre></div>
<div class="sourceCode" id="cb4"><pre class="sourceCode python"><code class="sourceCode python"><a class="sourceLine" id="cb4-1" data-line-number="1">prices <span class="op">=</span> pd.Series(x, index<span class="op">=</span>days)</a>
<a class="sourceLine" id="cb4-2" data-line-number="2">hypothetical <span class="op">=</span> prices.loc[<span class="st">&#39;21/5/2019&#39;</span>:]</a>
<a class="sourceLine" id="cb4-3" data-line-number="3">real <span class="op">=</span> prices.copy()</a>
<a class="sourceLine" id="cb4-4" data-line-number="4">real[<span class="st">&#39;21/5/2019&#39;</span>:] <span class="op">=</span> np.NaN</a></code></pre></div>
<div class="figure">
<div id="hypothetical-prices-figure" class="figure-frame">
<object class="svgplot " data="../slides/diagrams/data-science/hypothetical-prices.svg" width="80%" style=" ">
</object>
</div>
<div id="hypothetical-prices-magnify" class="magnify" onclick="magnifyFigure(&#39;hypothetical-prices&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="hypothetical-prices-caption" class="caption-frame">
<p>Figure: Anne has access to the share prices in the black stream but not in the blue stream. A hypothetical stream is the stream of future prices. Anne can define this hypothetical under constraints (latency, input etc). The need for a model is now exposed in the software infrastructure</p>
</div>
</div>
<h2 id="hypothetical-streams">Hypothetical Streams</h2>
<p>We’ll call the future price a hypothetical stream.</p>
<p>A hypothetical stream is a desired stream of information which cannot be directly accessed. The lack of direct access may be because the events happen in the future, or there may be some latency between the event and the availability of the data.</p>
<p>Any hypothetical stream will only be provided as a prediction, ideally with an error bar.</p>
<p>The nature of the hypothetical Anne needs is dependent on her decision-making process. In Anne’s case it will depend over what period she is expecting her returns. In MDOP Anne specifies a hypothetical that is derived from the pricing stream.</p>
<p>It is not the price stream directly, but Anne looks for <em>future</em> predictions from the price stream, perhaps for price in <span class="math inline"><em>T</em></span> days’ time.</p>
<p>At this stage, this stream is merely typed as a hypothetical.</p>
<p>There are constraints on the hypothetical, they include: the <em>input</em> information, the upper limit of latency between input and prediction, and the decision Anne needs to make (how far ahead, what her upside, downside risks are). These three constraints mean that we can only recover an approximation to the hypothetical.</p>
<h2 id="hypothetical-advantage">Hypothetical Advantage</h2>
<p>What is the advantage to defining things in this way? By defining, clearly, the two streams as real and hypothetical variants of each other, we now enable automation of the deployment and any redeployment process. The hypothetical can be <em>instantiated</em> against the real, and design criteria can be constantly evaluated triggering retraining when necessary.</p>
<h2 id="safeboda-edit">SafeBoda <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_ai/includes/safe-boda.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_ai/includes/safe-boda.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="safe-boda-system-figure" class="figure-frame">
<div class="centered" style="">
<img class="" src="../slides/diagrams/ai/safe-boda.png" width="60%" height="auto" align="center" style="background:none; border:none; box-shadow:none; display:block; margin-left:auto; margin-right:auto;vertical-align:middle">
</div>
</div>
<div id="safe-boda-system-magnify" class="magnify" onclick="magnifyFigure(&#39;safe-boda-system&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="safe-boda-system-caption" class="caption-frame">
<p>Figure: SafeBoda is a ride allocation system for Boda Boda drivers. Let’s imagine the capabilities we need for such an AI system.</p>
</div>
</div>
<p><a href="https://safeboda.com/ug/index.php#whysafeboda">SafeBoda</a> is a Kampala based rider allocation system for Boda Boda drivers. Boda boda are motorcycle taxis which give employment to, often young men, across Kampala. Safe Boda is driven by the knowledge that road accidents are set to match HIV/AIDS as the highest cause of death in low/middle income families by 2030.</p>
<blockquote>
<p>With road accidents set to match HIV/AIDS as the highest cause of death in low/middle income countries by 2030, SafeBoda’s aim is to modernise informal transportation and ensure safe access to mobility.</p>
</blockquote>
<div class="figure">
<div id="ride-allocation-system-figure" class="figure-frame">
<object class="svgplot " data="../slides/diagrams/ai/ride-allocation-prediction.svg" width="60%" style=" ">
</object>
</div>
<div id="ride-allocation-system-magnify" class="magnify" onclick="magnifyFigure(&#39;ride-allocation-system&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="ride-allocation-system-caption" class="caption-frame">
<p>Figure: Some software components in a ride allocation system. Circled components are hypothetical, rectangles represent actual data.</p>
</div>
</div>
<p>Let’s consider a ride sharing app, for example the SafeBoda system.</p>
<p>Anne is on her way home now; she wishes to hail a car using a ride sharing app.</p>
<p>The app is designed in the following way. On opening her app Anne is notified about drivers in the nearby neighborhood. She is given an estimate of the time a ride may take to come.</p>
<p>Given this information about driver availability, Anne may feel encouraged to enter a destination. Given this destination, a price estimate can be given. This price is conditioned on other riders that may wish to go in the same direction, but the price estimate needs to be made before the user agrees to the ride.</p>
<p>Business customer service constraints dictate that this price may not change after Anne’s order is confirmed.</p>
<p>In this simple system, several decisions are being made, each of them on the basis of a hypothetical.</p>
<p>When Anne calls for a ride, she is provided with an estimate based on the expected time a ride can be with her. But this estimate is made without knowing where Anne wants to go. There are constraints on drivers imposed by regional boundaries, reaching the end of their shift, or their current passengers mean that this estimate can only be a best guess.</p>
<p>This best guess may well be driven by previous data.</p>
<h2 id="ride-sharing-service-oriented-to-data-oriented-edit">Ride Sharing: Service Oriented to Data Oriented <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/ride-sharing-soa-doa.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/ride-sharing-soa-doa.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<div class="figure">
<div id="ride-share-service-soa-figure" class="figure-frame">
<object class="svgplot " data="../slides/diagrams/data-science/ride-share-service-soa.svg" width="80%" style=" ">
</object>
</div>
<div id="ride-share-service-soa-magnify" class="magnify" onclick="magnifyFigure(&#39;ride-share-service-soa&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="ride-share-service-soa-caption" class="caption-frame">
<p>Figure: Service oriented architecture. The data access is buried in the cost allocation service. Data dependencies of the service cannot be found without trawling through the underlying code base.</p>
</div>
</div>
<p>The modern approach to software systems design is known as a <em>service-oriented architectures</em> (SOA). The idea is that software engineers are responsible for the availability and reliability of the API that accesses the service they own. Quality of service is maintained by rigorous standards around <em>testing</em> of software systems.</p>
<div class="figure">
<div id="ride-share-service-doa-figure" class="figure-frame">
<object class="svgplot " data="../slides/diagrams/data-science/ride-share-service-doa.svg" width="80%" style=" ">
</object>
</div>
<div id="ride-share-service-doa-magnify" class="magnify" onclick="magnifyFigure(&#39;ride-share-service-doa&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="ride-share-service-doa-caption" class="caption-frame">
<p>Figure: Data oriented architecture. Now the joins and the updates are exposed within the streaming ecosystem. We can programatically determine the factor graph which gives the thread through the model.</p>
</div>
</div>
<p>In data driven decision-making systems, the quality of decision-making is determined by the quality of the data. We need to extend the notion of <em>service</em>-oriented architecture to <em>data</em>-oriented architecture (DOA).</p>
<p>The focus in SOA is eliminating <em>hard</em> failures. Hard failures can occur due to bugs or systems overload. This notion needs to be extended in ML systems to capture <em>soft failures</em> associated with declining data quality, incorrect modeling assumptions and inappropriate re-deployments of models. We need to focus on data quality assessments. In data-oriented architectures engineering teams are responsible for the <em>quality</em> of their output data streams in addition to the <em>availability</em> of the service they support <span class="citation" data-cites="Lawrence:drl17">(Lawrence 2017)</span>. Quality here is not just accuracy, but fairness and explainability. This important cultural change would be capable of addressing both the challenge of <em>technical debt</em> <span class="citation" data-cites="Sculley:debt15">(Sculley et al. 2015)</span> and the social responsibility of ML systems.</p>
<p>Software development proceeds with a <em>test-oriented</em> culture. One where tests are written before software, and software is not incorporated in the wider system until all tests pass. We must apply the same standards of care to our ML systems, although for ML we need statistical tests for quality, fairness and consistency within the environment. Fortunately, the main burden of this testing need not fall to the engineers themselves: through leveraging <em>classical statistics</em> and <em>emulation</em> we will automate the creation and redeployment of these tests across the software ecosystem, we call this <em>ML hypervision</em> (WP5 ).</p>
<p>Modern AI can be based on ML models with many millions of parameters, trained on very large data sets. In ML, strong emphasis is placed on <em>predictive accuracy</em> whereas sister-fields such as statistics have a strong emphasis on <em>interpretability</em>. ML models are said to be ‘black boxes’ which make decisions that are not explainable.<a href="#fn3" class="footnote-ref" id="fnref3"><sup>3</sup></a></p>
<div class="figure">
<div id="ride-share-service-doa-hypothetical-figure" class="figure-frame">
<object class="svgplot " data="../slides/diagrams/data-science/ride-share-service-doa-hypothetical.svg" width="80%" style=" ">
</object>
</div>
<div id="ride-share-service-doa-hypothetical-magnify" class="magnify" onclick="magnifyFigure(&#39;ride-share-service-doa-hypothetical&#39;)">
<p><img class="img-button" src="{{ "/assets/images/Magnify_Large.svg" | relative_url }}" style="width:1.5ex"></p>
</div>
<div id="ride-share-service-doa-hypothetical-caption" class="caption-frame">
<p>Figure: Data-oriented programing. There is a requirement for an estimate of the driver allocation to give a rough cost estimate before the user has confirmed the ride. In data-oriented programming, this is achieved through declaring a hypothetical stream which approximates the true driver allocation, but with restricted input information and constraints on the computational latency.</p>
</div>
</div>
<p>For the ride sharing system, we start to see a common issue with a more complex algorithmic decision-making system. Several decisions are being made multilple times. Let’s look at the decisions we need along with some design criteria.</p>
<ol type="1">
<li>Driver Availability: Estimate time to arrival for Anne’s ride using Anne’s location and local available car locations. Latency 50 milliseconds</li>
<li>Cost Estimate: Estimate cost for journey using Anne’s destination, location and local available car current destinations and availability. Latency 50 milliseconds</li>
<li>Driver Allocation: Allocate car to minimize transport cost to destination. Latency 2 seconds.</li>
</ol>
<p>So we need:</p>
<ol type="1">
<li>a hypothetical to estimate availability. It is constrained by lacking destination information and a low latency requirement.</li>
<li>a hypothetical to estimate cost. It is constrained by low latency requirement and</li>
</ol>
<p>Simultaneously, drivers in this data ecosystem have an app which notifies them about new jobs and recommends them where to go.</p>
<p>Further advantages. Strategies for data retention (when to snapshot) can be set globally.</p>
<p>A few decisions need to be made in this system. First of all, when the user opens the app, the estimate of the time to the nearest ride may need to be computed quickly, to avoid latency in the service.</p>
<p>This may require a quick estimate of the ride availability.</p>
<h2 id="information-dynamics-edit">Information Dynamics <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/information-dynamics.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/information-dynamics.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>With all the second guessing within a complex automated decision-making system, there are potential problems with information dynamics, the ‘closed loop’ problem, where the sub-systems are being approximated (second guessing) and predictions downstream are being affected.</p>
<p>This leads to the need for a closed loop analysis, for example, see the <a href="https://www.gla.ac.uk/schools/computing/research/researchsections/ida-section/closedloop/">“Closed Loop Data Science”</a> project led by Rod Murray-Smith at Glasgow.</p>
<h2 id="conclusion-edit">Conclusion <span class="editsection-bracket" style="">[</span><span class="editsection" style=""><a href="https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-oriented-conclusions.md" target="_blank" onclick="ga('send', 'event', 'Edit Page', 'Edit', 'https://github.com/lawrennd/snippets/edit/main/_data-science/includes/data-oriented-conclusions.md', 13);">edit</a></span><span class="editsection-bracket" style="">]</span></h2>
<p>We operate in a technologically evolving environment. Machine learning is becoming a key coponent in our decision-making capabilities, our intelligence and strategic command. However, technology drove changes in battlefield strategy. From the stalemate of the first world war to the tank-dominated Blitzkrieg of the second, to the asymmetric warfare of the present. Our technology, tactics and strategies are also constantly evolving. Machine learning is part of that evolution solution, but the main challenge is not to become so fixated on the tactics of today that we miss the evolution of strategy that the technology is suggesting.</p>
<p>Data oriented programming offers a set of development methodologies which ensure that the system designer considers what decisions are required, how they will be made, and critically, declares this within the system architecture.</p>
<p>This allows for monitoring of <em>data quality</em>, <em>fairness</em>, <em>model accuracy</em> and opens the door to Auto AI: a more sophisticated form of auto ML where full redployments of models are considered while analyzing the information dynamics of a complex automated decision-making system.</p>
<h1 id="references" class="unnumbered">References</h1>
<div id="refs" class="references">
<div id="ref-Cooper:transformation91">
<p>Cooper, Brian. 1991. <em>Transformation of a Valley: Derbyshire Derwent</em>. Scarthin Books.</p>
</div>
<div id="ref-Lawrence:drl17">
<p>Lawrence, Neil D. 2017. “Data Readiness Levels.” arXiv.</p>
</div>
<div id="ref-Maxwell:governors1867">
<p>Maxwell, James Clerk. 1867. “On Governors.” <em>Proceedings of the Royal Society of London</em> 16. The Royal Society: 270–83. <a href="http://www.jstor.org/stable/112510" class="uri">http://www.jstor.org/stable/112510</a>.</p>
</div>
<div id="ref-Sculley:debt15">
<p>Sculley, D., Gary Holt, Daniel Golovin, Eugene Davydov, Todd Phillips, Dietmar Ebner, Vinay Chaudhary, Michael Young, Jean-François Crespo, and Dan Dennison. 2015. “Hidden Technical Debt in Machine Learning Systems.” In <em>Advances in Neural Information Processing Systems 28</em>, edited by Corinna Cortes, Neil D. Lawrence, Daniel D. Lee, Masashi Sugiyama, and Roman Garnett, 2503–11. Curran Associates, Inc. <a href="http://papers.nips.cc/paper/5656-hidden-technical-debt-in-machine-learning-systems.pdf" class="uri">http://papers.nips.cc/paper/5656-hidden-technical-debt-in-machine-learning-systems.pdf</a>.</p>
</div>
<div id="ref-Wiener:cybernetics48">
<p>Wiener, Norbert. 1948. <em>Cybernetics: Control and Communication in the Animal and the Machine</em>. Cambridge, MA: MIT Press.</p>
</div>
</div>
<section class="footnotes">
<hr />
<ol>
<li id="fn1"><p>This is related to challenges of machine learning and technical debt <span class="citation" data-cites="Sculley:debt15">(Sculley et al. 2015)</span>, although we are trying to frame the solution here rather than the problem.<a href="#fnref1" class="footnote-back">↩</a></p></li>
<li id="fn2"><p>These approaches are one area of focus for my own team’s research. A data first architecture is a prerequisite for efficient deployment of machine learning systems.<a href="#fnref2" class="footnote-back">↩</a></p></li>
<li id="fn3"><p>See for example <a href="https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/">“The Dark Secret at the Heart of AI” in Technology Review</a>.<a href="#fnref3" class="footnote-back">↩</a></p></li>
</ol>
</section>