Adaptive Network Sparsification with Dependent Variational Beta-Bernoulli Dropout
This work addresses the challenge of sparsifying neural networks without accuracy loss for machine learning practitioners, though it is incremental as it builds on existing variational dropout approaches.
The paper tackles the problem of suboptimal network sparsification in variational dropout by proposing an adaptive method that sets dropout rates based on input data, allowing neurons to be generic, specific, or dropped. It results in more compact networks with consistent accuracy improvements on multiple public datasets.
While variational dropout approaches have been shown to be effective for network sparsification, they are still suboptimal in the sense that they set the dropout rate for each neuron without consideration of the input data. With such input-independent dropout, each neuron is evolved to be generic across inputs, which makes it difficult to sparsify networks without accuracy loss. To overcome this limitation, we propose adaptive variational dropout whose probabilities are drawn from sparsity-inducing beta Bernoulli prior. It allows each neuron to be evolved either to be generic or specific for certain inputs, or dropped altogether. Such input-adaptive sparsity-inducing dropout allows the resulting network to tolerate larger degree of sparsity without losing its expressive power by removing redundancies among features. We validate our dependent variational beta-Bernoulli dropout on multiple public datasets, on which it obtains significantly more compact networks than baseline methods, with consistent accuracy improvements over the base networks.