Contrastive predictive coding - NLP

• This repository contains an implementation of the NLP experiment conducted in the original CPC paper (2018)
• This project was conducted for the Machine Learning II course, which is part of the first year of the Master's degree program in Computer Science at the University of Mons
• Link to the original paper

Results

• The model has been trained on a limited subset of the bookcorpus dataset due to computational constraints
• The sentence encoder, with both random and trained weights, was evaluated using the TREC dataset with coarse labels

Train results

Loss	Accuracy

TREC results

model_name	train_accuracy	test_accuracy
random_encoder	0.387931	0.302
trained_encoder	0.782465	0.772

Usage

• Run python main.py
• See configuration files in configs directory