A Contrastive Learning Approach for Training Variational Autoencoder Priors
Summary of Dai et al.'s NeurIPS 2021 paper introducing Noise Contrastive Priors (NCPs), a novel method using noise contrastive estimation to train …
Representation-Learning
2 posts tagged with representation-learning.
EigenNoise: A Contrastive Prior to Warm-Start Representations
Novel initialization scheme for word vectors based on dense co-occurrence modeling, achieving competitive performance with GloVe without requiring …