Notebooks from the Seminars:

Introduction to Artificial Intelligence Programming < WS18/19 - WS19/20

Georg Trogemann und Christian Heck

Grundlagenseminar Material/Skulptur/Code

Dienstag wöchentlich 11:00 –13:00

Filzengraben 8 - 10, 0.2 Experimentelle Informatik

Kunsthochschule für Medien Köln

Description

Profound cultural consequences of AI do not only appear with the use of upload filters for algorithmic censorship of undesirable text and image content or the auctioning of AI paintings at Christie's; nor with the formulation of ethical guidelines for dealing with AI or the increased emergence of AI-powered hate speech bots. They begin, quite abstractly and mostly unnoticed in their programming, in semi-public, very formal fields of discourse.

This is exactly where we start experimentally. The seminar provides a very elementary introduction to the subsymbolic AI of neural networks and their programming. The aim of this seminar is to code from scratch, discuss the code together and learn to understand it, in order to learn to assess the possibilities, limits and dangers of this technology for oneself.

We do not adopt the technology of artificial intelligence as a tool in the Homo Faberian sense, but combine programming as an artistic practice with the critical analysis of its social effects.

Info

Seminar Wiki Pages:

• WS18-19
• WS19-20

Executing the Notebooks:

• You can run, execute and work on the following Notebooks here:

Setting Up

Basics in Anaconda & Jupyter Notebooks:

• KI_Workaround_installieren

---

Hands on Python

see repository: https://github.com/experimental-informatics/hands-on-python

---

Artificial Neural Net in Python

many of the Code is based on Tariq Rashid's Book »Neuronale Netze selbst programmieren« < Git Repo

• Coding a Dense Neural Net in Python from Scratch:

  • KNN_in-Python_am-eigenen-Bild.ipynb

• Data Preprocessing:

  • Preprocessing.ipynb

---

Hands on Keras & Tensorflow

all examples working with MNIST handwritten digit database

most of the Codes are based on Francois Chollet's Book »Deep Learning with Python« < Git Repo

Dense Neural Net

• Coding aArtificial Neural Net (DNN) in Tensorflow & Keras:
  • ANN-in-Keras.ipynb

Autoencoder

• Simple Autoencoder in Tensorflow & Keras
  • Autoencoder-in-Keras.ipynb

Convolutional Neural Net

• CNN in Tensorflow & Keras
  • CNN-in-Keras.ipynb

Generative Adversarial Network

• GAN in Tensorflow & Keras
  • GAN-in-Keras.ipynb

Interpretable AI

• Visualize Activations (based on model from CNN-in-Keras.ipynb)

  • mnist_cnn-activations_keract.ipynb

• Visualize Heatmaps (based on model from CNN-in-Keras.ipynb)

  • mnist_cnn-heatmaps_keract.ipynb

Explainable AI

• LIME for image classification by using Keras (InceptionV3)
  • kiss-explained_lime-demo.ipynb

---

Natural Language Processing (NLP)

Text Preprocessing

• for english Textcorpora
  • tokenize_english_textcorpora.ipynb
• for german Textcorpora
  • tokenize_german_textcorpora.ipynb

Chatbots

• Basic Encodings & traditional embeddings (ONE-HOT / BOW / TF-IDF)
  • one-hot_bow_tfidf.ipynb
• Basic Chatbots (TF-IDF)
  • simple_german_chatbot.ipynb
  • simple_english_chatbot.ipynb
• Chatbots with Chatterbot
  • chatterbot.ipynb

Sentiment Analysis

• Sentiment Analysis for german textcorpora
  • Sentiment-analyses_4_german-texts_with_TextBlob.ipynb

Word embeddings

• train a Word2vev Modell on your own Texts
  • train_your_own_Word2vec-model.ipynb

RNN/LSTM

• LSTM- Textgeneration
  • LSTM-Textgenerator.ipynb

Datasets

• Loading and scraping data from the web
  • load_scrape_data-sets.ipynb

---