Submodular Batch Selection for Training Deep Neural Networks
This addresses training efficiency for deep learning practitioners, but it is incremental as it builds on existing batch selection methods.
The paper tackled the problem of improving training efficiency and generalization in deep neural networks by introducing a mini-batch selection strategy based on submodular function maximization, which captures sample informativeness and diversity, and experiments showed it provides better generalization than Stochastic Gradient Descent and a baseline across various settings.
Mini-batch gradient descent based methods are the de facto algorithms for training neural network architectures today. We introduce a mini-batch selection strategy based on submodular function maximization. Our novel submodular formulation captures the informativeness of each sample and diversity of the whole subset. We design an efficient, greedy algorithm which can give high-quality solutions to this NP-hard combinatorial optimization problem. Our extensive experiments on standard datasets show that the deep models trained using the proposed batch selection strategy provide better generalization than Stochastic Gradient Descent as well as a popular baseline sampling strategy across different learning rates, batch sizes, and distance metrics.