A Simple Approach To Define Curricula For Training Neural Networks

In practice, sequence of mini-batches generated by uniform sampling of examples from the entire data is used for training neural networks. Curriculum learning is a training strategy that sorts the training examples by their difficulty and gradually exposes them to the learner. In this work, we propose two novel curriculum learning algorithms and empirically show their improvements in performance with convolutional and fully-connected neural networks on multiple datasets. Our dynamic curriculum learning algorithm tries to reduce the distance between the network weight and the optimal weight at any training step by greedily sampling examples with gradients that are directed towards the optimal weight. It achieves a speedup of $\sim 45\%$ in our experiments. We also introduce a new task-specific curriculum learning strategy that uses statistical measures such as standard deviation and entropy values to score the difficulty of data points in natural image datasets. We show that this new approach yields a mean speedup of $\sim 43\%$ in the experiments we perform. Further, we also use our algorithms to learn why curriculum learning works. Based on our study, we argue that curriculum learning removes noisy examples from the initial phases of training, and gradually exposes them to the learner acting like a regularizer that helps in improving the generalization ability of the learner.