interpreting-rl-behavior interpreting-rl-behavior.github.io issues

interpreting-rl-behavior / interpreting-rl-behavior.github.io

Code for the site https://interpreting-rl-behavior.github.io/

Creative Commons Attribution 4.0 International

0 stars 0 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Add tooltips to panel

#78 danbraunai opened 2 years ago
1
Setting filters for a specific set of timesteps

#77 leesharkey closed 2 years ago
1
Filters don't reset in the dropdown menu despite clicking reset

#76 leesharkey closed 2 years ago
1
Write the methods section

#75 leesharkey opened 2 years ago
0
Write clustering algorithm for hx/hx_grad/env_grad data to improve sample filtering

#74 leesharkey closed 2 years ago
1
Add plot titles/labels to the action/value readout matrix/bias plots

#73 leesharkey closed 2 years ago
1
Update xcaus plotting code so that we can create plots for sets of samples that have been filtered according to hx_activations

#72 leesharkey closed 2 years ago
1
Add legend to saliency plot

#71 danbraunai closed 2 years ago
4
Use different colours in scatterplot

#70 danbraunai closed 2 years ago
0
ICA scatter plot and Bar chart need axis labels (like 'IC loadings' and 'IC dimensions')

#69 danbraunai closed 2 years ago
1
Remove unsued code templates

#68 danbraunai opened 2 years ago
0
Identify stories using dataset examples for all directions

#67 leesharkey closed 2 years ago
0
Handle extreme points and bars in panel

#66 danbraunai closed 2 years ago
2
Remedy the unexpected broadcasting behaviour for extrema-finding code in import_data.py

#65 leesharkey closed 2 years ago
2
Dataset examples identifier script

#64 leesharkey closed 2 years ago
1
Allow maximisation of each chart in panel

#63 danbraunai opened 2 years ago
1
Fix html error with npm run build

#62 danbraunai closed 2 years ago
0
Get all required target functions working

#61 leesharkey closed 2 years ago
2
Identify stories in the ICA gradients in the panel

#60 leesharkey closed 2 years ago
1
Add indicator of the agent's action at each timestep

#59 danbraunai closed 2 years ago
2
Add saliency timestep indicator

#58 danbraunai closed 2 years ago
3
Convert saliency type to dropdown if more then 2 options

#57 danbraunai closed 2 years ago
0
Make the left-right arrows into buttons with round borders

#56 danbraunai closed 2 years ago
0
Customise the samples and saliency type shown in each panel

#55 danbraunai closed 2 years ago
5
Implement saving of config file for every script.

#54 leesharkey closed 2 years ago
0
Fix losses. There is something fishing going on.

#53 leesharkey closed 2 years ago
2
Begin putting article text, figure, references into ICML format

#52 leesharkey closed 2 years ago
2
Implement ICA gradient projection and modify panel data and text so that it takes generic 'components'

#51 leesharkey closed 2 years ago
1
Figure out why the agent always goes down-left when the screen is black

#50 leesharkey closed 2 years ago
1
Swap the ResNets in the encoder/initializer to be smaller convnets like in the main RSSm

#49 leesharkey closed 2 years ago
1
Fix saliency experiments for new generative model

#48 leesharkey closed 2 years ago
1
Adapt the loss_over_time plot for the new generative model

#47 leesharkey closed 2 years ago
1
Make a script called record_gen_model.py that does the job of record_informinit_gen_model.py and record_random_gen_model.py

#46 leesharkey closed 2 years ago
1
Design experiments that more comprehensively demonstrates that the environment simulator approximates the causality of the real environment

#45 leesharkey closed 2 years ago
2
Address the issue of correlated early gradients in the article text

#44 leesharkey opened 2 years ago
1
Change MDP diagram label "Memory; M_t" so that it's "Agent hidden state; h_t"

#43 leesharkey closed 2 years ago
0
Create new architecture diagram in article

#42 leesharkey closed 2 years ago
1
Reimplement the VAE encoder and VAE latent loss

#41 leesharkey closed 2 years ago
1
Identify why the image never goes black when using both sim actions and sim images

#40 leesharkey closed 2 years ago
2
Confirm that actions, agent hx, and observations are aligned in the gen model

#39 leesharkey closed 2 years ago
2
Do the maths for using ICA components instead of PCA components

#38 leesharkey closed 2 years ago
1
Once main interpretation is done, make a decision on whether to do 'adversarial environment' validation experiments

#37 leesharkey closed 2 years ago
2
Retrain gen model, demonstrate independence of agent and env, then resume interpreting.

#36 leesharkey closed 2 years ago
1
Concatenate all images from a sequence to one file

#35 danbraunai closed 2 years ago
7
Parallelize record.py

#34 leesharkey closed 2 years ago
1
Fix dataloader for gen model

#33 leesharkey closed 2 years ago
3
Setup ids for clickable table of contents

#32 danbraunai opened 2 years ago
0
Tweak architecture diagram colours to conform to MDP diagram

#31 leesharkey opened 2 years ago
0
Grad_hx_hx_direction_i_loadings all generate identically

#30 NixGD closed 2 years ago
2
Gradient of hx at timestep 0 of sample 00000 highly correlated

#29 leesharkey closed 2 years ago
3