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:
- MNIST
- Morpho-MNIST
- CIFAR-10
- Fetal Ultrasound Images (Sononet classifier based on code by Tianrui Liu)