Incremental Sequence Learning

Deep learning research over the past years has shown that by increasing the scope or difficulty of the learning problem over time, increasingly complex learning problems can be addressed. We study incremental learning in the context of sequence learning, using generative RNNs in the form of multi-layer recurrent Mixture Density Networks. While the potential of incremental or curriculum learning to enhance learning is known, indiscriminate application of the principle does not necessarily lead to improvement, and it is essential therefore to know which forms of incremental or curriculum learning have a positive effect. This research contributes to that aim by comparing three instantiations of incremental or curriculum learning. We introduce Incremental Sequence Learning, a simple incremental approach to sequence learning. Incremental Sequence Learning starts out by using only the first few steps of each sequence as training data. Each time a performance criterion has been reached, the length of the parts of the sequences used for training is increased. To evaluate Incremental Sequence Learning and comparison methods, we introduce and make available a novel sequence learning task and data set: predicting and classifying MNIST pen stroke sequences, where the familiar handwritten digit images have been transformed to pen stroke sequences representing the skeletons of the digits. We find that Incremental Sequence Learning greatly speeds up sequence learning and reaches the best test performance level of regular sequence learning 20 times faster, reduces the test error by 74%, and in general performs more robustly; it displays lower variance and achieves sustained progress after all three comparison methods have stopped improving. The two other instantiations of curriculum learning do not result in any noticeable improvement. A trained sequence prediction model is also used in transfer learning to the task of sequence classification, where it is found that transfer learning realizes improved classification performance compared to methods that learn to classify from scratch.