[DL輪読会]Recent Advances in Autoencoder-Based Representation Learning 1. Recent advances in autoencoder-based representation learning include incorporating meta-priors to encourage disentanglement and using rate-distortion and rate-distortion-usefulness tradeoffs to balance compression and reconstruction. 2. Variational autoencoders introduce priors to disentangle latent factors, but recent work agg