/
Visualization.Rmd
71 lines (41 loc) · 1.1 KB
/
Visualization.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
---
title: 'Visualization: The Plots Thicken'
author: "Augustina Ragwitz"
output: html_document
params:
data_folder: "downloads"
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
```{r includes}
library(here)
library(keras)
library(tidyverse)
library(lime)
```
# Mission
Explore ways of natively visualizing models in R!
* Visualize layer content and structure
* Generate heatmaps to see how the convnet learned
# Generate a Heatmap to see how the Convnet Learned
## Use a 2D Convnent
Let's use a 2d convnent for MNIST instead of dense layers.
```{r}
mnist <- dataset_mnist()
train_images <- mnist$train$x
train_labels <- mnist$train$y
# TODO improve
# TODO show adding a layer from another model
```
# Show what parameters impacted the prediction with LIME
# Visualize Layer and Model Content
## Greta
Greta is a package to do Bayesian statistics via TensorFlow
https://greta-dev.github.io/greta/
https://cran.r-project.org/web/packages/greta/vignettes/get_started.html
## Diagrammer
http://rich-iannone.github.io/DiagrammeR/
Or Greta
## iGraph
I tried and failed :(