Projective Latent Interventions

Projective Latent Interventions are a technique to inform the learning of neural network classifiers by manually manipulating embedded latent spaces.

Paper preprint available on arXiv.

Requires this PyTorch implementation of parametric t-SNE.

This repository contains application examples for the following datasets:

Name		Name	Last commit message	Last commit date
Embeddings		Embeddings
Experiments/mnist		Experiments/mnist
checkpoints/ifind1_sononet_8_6classes		checkpoints/ifind1_sononet_8_6classes
configs		configs
functions		functions
logs		logs
sononet		sononet
torchsample		torchsample
.gitignore		.gitignore
Automated Experiments.ipynb		Automated Experiments.ipynb
README.md		README.md
Untitled.ipynb		Untitled.ipynb
Untitled1.ipynb		Untitled1.ipynb
fit_intervention.py		fit_intervention.py
plsd-cifar-Gini.ipynb		plsd-cifar-Gini.ipynb
plsd-cifar-umap-Gini.ipynb		plsd-cifar-umap-Gini.ipynb
plsd-cifar.ipynb		plsd-cifar.ipynb
plsd-mnist-Gini.ipynb		plsd-mnist-Gini.ipynb
plsd-mnist-ptsne-keras-Gini.ipynb		plsd-mnist-ptsne-keras-Gini.ipynb
plsd-mnist-umap-Gini.ipynb		plsd-mnist-umap-Gini.ipynb
plsd-mnist.ipynb		plsd-mnist.ipynb
plsd-morpho-mnist.ipynb		plsd-morpho-mnist.ipynb
plsd-ultrasound-Gini.ipynb		plsd-ultrasound-Gini.ipynb
plsd-ultrasound.ipynb		plsd-ultrasound.ipynb
prototype-export.ipynb		prototype-export.ipynb

Navigation Menu