2 papers
cs.CV2011★ 8 cited
Learning Representations by Maximizing Compression
Karol Gregor, Yann LeCun
We give an algorithm that learns a representation of data through compression. The algorithm 1) predicts bits sequentially from those previously seen and 2) has a structure and a n…
cs.CV2011★ 5 cited
Efficient Learning of Sparse Invariant Representations
Karol Gregor, Yann LeCun
We propose a simple and efficient algorithm for learning sparse invariant representations from unlabeled data with fast inference. When trained on short movies sequences, the learn…