Determinantal Point Processes for Mini-Batch Diversification
arXiv:1705.00607
Abstract
We study a mini-batch diversification scheme for stochastic gradient descent (SGD). While classical SGD relies on uniformly sampling data points to form a mini-batch, we propose a non-uniform sampling scheme based on the Determinantal Point Process (DPP). The DPP relies on a similarity measure between data points and gives low probabilities to mini-batches which contain redundant data, and higher probabilities to mini-batches with more diverse data. This simultaneously balances the data and leads to stochastic gradients with lower variance. We term this approach Diversified Mini-Batch SGD (DM-SGD). We show that regular SGD and a biased version of stratified sampling emerge as special cases. Furthermore, DM-SGD generalizes stratified sampling to cases where no discrete features exist to bin the data into groups. We show experimentally that our method results more interpretable and diverse features in unsupervised setups, and in better classification accuracies in supervised setups.
References in corpus (12)
- Adam: A Method for Stochastic Optimization
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- ADADELTA: An Adaptive Learning Rate Method
- CNN Features off-the-shelf: an Astounding Baseline for Recognition
- Determinantal point processes for machine learning
- Variational Bayesian Inference with Stochastic Search
- Accelerating Minibatch Stochastic Gradient Descent using Stratified Sampling
- Fast DPP Sampling for Nyström with Application to Kernel Methods
- Efficient Sampling for k-Determinantal Point Processes
- Importance Sampling for Minibatches
- Variational Tempering
- On Using Control Variates with Stochastic Approximation for Variational Bayes and its Connection to Stochastic Linear Regression