Please join us for today's tea-talk. Caglar will be doing a practice talk for his upcoming workshop poster at the NIPS Deep Learning Workshop. Time-willing, I will then present the paper entitled "Cardinality Restricted Boltzmann Machines" by Kevin Swersky et al. We may also have special appearances by Ian Goodfellow and Razvan Pascanu.
Abstract (Caglar Gulcehre):
We explore the effect of introducing prior information into the intermediate level of deep learning algorithms for a learning task on which all the state-of-the-art machine learning algorithms tested failed to learn. We motivate our work from the hypothesis that humans learn such intermediate concepts from other individuals via a form of supervision or guidance using a curriculum. The experiments we have conducted provide positive evidence in favor of this hypothesis. In our experiments, a two-tiered MLP architecture is trained on a dataset with 64x64 binary inputs images, each image with three sprites. The final task is to decide whether all the sprites are the same or one of them is different. Sprites are pentomino tetris shapes and they are placed in an image with different locations using scaling and rotation transformations. The first level of the two-tiered MLP is pretrained with intermediate level targets being the presence of sprites at each location, while the second level takes the output of the first level as input and predicts the final task target binary event. The two-tiered MLP architecture, with a few tens of thousand examples, was able to learn the task perfectly, whereas all other algorithms (include unsupervised pre-training, but also traditional algorithms like SVMs, decision trees and boosting) all perform no better than chance. We hypothesize that the optimization difficulty involved when the intermediate pre-training is not performed is due to the composition of two highly non-linear tasks. Our findings are also consistent with hypotheses on cultural learning inspired by the observations of optimization problems with deep learning, presumably because of effective local minima.