Mattéo Delabre
|
73d2572fbe
|
Fix typo in README
|
3 years ago |
Mattéo Delabre
|
6921be6ca6
|
Augment illustration
|
3 years ago |
Mattéo Delabre
|
54813c4c46
|
Fixed name clash in plot.py
|
3 years ago |
Mattéo Delabre
|
376c7c0bc4
|
Add illustration
|
3 years ago |
Mattéo Delabre
|
cacf4408e5
|
Add README
|
3 years ago |
Mattéo Delabre
|
af8862d360
|
Reflect parameter changes in `experiment.py`
|
3 years ago |
Mattéo Delabre
|
d8a140d793
|
Simplify arguments for ConstellationNet
|
3 years ago |
Mattéo Delabre
|
a7e9dd2230
|
Save whole model to avoid definition duplicate
|
3 years ago |
Mattéo Delabre
|
0769a61fcf
|
Use best parameters as found by experimentation
|
3 years ago |
Mattéo Delabre
|
5cb087d971
|
Add experiment results
|
3 years ago |
Mattéo Delabre
|
fe23fda86d
|
Shuffle configurations in experiment
|
3 years ago |
Mattéo Delabre
|
8d64f91661
|
Add partitioning to experiment
|
3 years ago |
Mattéo Delabre
|
0dfbed1bb7
|
Add experiment code
|
3 years ago |
Mattéo Delabre
|
7efcbd3948
|
Add product_dict util function
|
3 years ago |
Mattéo Delabre
|
e8e4dcec83
|
Set learning rate to 10^-3
|
3 years ago |
Mattéo Delabre
|
af59421dc0
|
Fix variable naming clash
|
3 years ago |
Mattéo Delabre
|
fb2518b321
|
Show final loss after training
|
3 years ago |
Mattéo Delabre
|
1d39184036
|
Use euclidean distance for stop criterion
|
3 years ago |
Mattéo Delabre
|
4f1514accf
|
Add noise to plot
|
3 years ago |
Mattéo Delabre
|
2c4786bdba
|
Remove reconstruction examples after training
|
3 years ago |
Mattéo Delabre
|
4bf0c0f363
|
Add batch normalization and ReLU activation
|
3 years ago |
Mattéo Delabre
|
8fa6b46ca8
|
Change training strategy to adaptive learning rate
|
3 years ago |
Mattéo Delabre
|
3f2c6d18a3
|
Manually seed training for reproducibility
|
3 years ago |
Mattéo Delabre
|
4d8d419d31
|
Only plot noise if noise_samples > 0
|
3 years ago |
Mattéo Delabre
|
2b93a5f1bc
|
Fix wrong input dimension for decision region plot
|
3 years ago |
Mattéo Delabre
|
e4457400a6
|
Add larger set of colors for plot
|
3 years ago |
Mattéo Delabre
|
722c02ef24
|
Cleanup channel model
|
3 years ago |
Mattéo Delabre
|
3638749998
|
Share plotting code between plot.py and train.py
|
3 years ago |
Mattéo Delabre
|
34cb3b863b
|
Simplify channel model to account for normalized power
|
3 years ago |
Mattéo Delabre
|
59d7adf6bd
|
Add convergence criterion plot during train
|
3 years ago |
Mattéo Delabre
|
3c199dfc41
|
Remove activation function from ConstellationNet
|
3 years ago |
Mattéo Delabre
|
989a51b72e
|
Increase batch size halfway through training
|
3 years ago |
Mattéo Delabre
|
3365603614
|
Add normalization layer
|
3 years ago |
Mattéo Delabre
|
b97ba61f42
|
Improve plot legibility
|
3 years ago |
Mattéo Delabre
|
8a31b22b83
|
Increase noise stddev
|
3 years ago |
Mattéo Delabre
|
04c2938b9e
|
Decrease alpha for noise plot
|
3 years ago |
Mattéo Delabre
|
9a87b322d4
|
Plot noise
|
3 years ago |
Mattéo Delabre
|
6ea0e653c1
|
Balance training examples
|
3 years ago |
Mattéo Delabre
|
99c96162c0
|
Use SELU instead of ReLU
|
3 years ago |
Mattéo Delabre
|
6d31ab3a13
|
Use ReLU with normalization
|
3 years ago |
Mattéo Delabre
|
197a01e993
|
Fix plot scaling
|
3 years ago |
Mattéo Delabre
|
8f6363ee21
|
Plot decision regions from the decoder
|
3 years ago |
Mattéo Delabre
|
3b40e27070
|
Add Gaussian channel model
|
3 years ago |
Mattéo Delabre
|
d0af8fc3da
|
Use Tanh activation to preserve negative values
|
3 years ago |
Mattéo Delabre
|
a01d83f339
|
Improve plot
|
3 years ago |
Mattéo Delabre
|
49e63775dd
|
Save trained models and plot encoding
|
3 years ago |
Mattéo Delabre
|
2af8354a07
|
Add initial implementation
|
3 years ago |
Mattéo Delabre
|
7463e9fe5b
|
Initial commit
|
3 years ago |