DeepHoyer: Learning Sparser Neural Network with Differentiable Scale-Invariant Sparsity Measures

Abstract

In seeking for sparse and efficient neural network models, many previous works investigated on enforcing L1 or L0 regularizers to encourage weight sparsity during training. The L0 regularizer measures the parameter sparsity directly and is invariant to the scaling of parameter values, but it cannot provide useful gradients, and therefore requires complex optimization techniques. The L1 regularizer is almost everywhere differentiable and can be easily optimized with gradient descent. Yet it is not scale-invariant, causing the same shrinking rate to all parameters, which is inefficient in increasing sparsity. Inspired by the Hoyer measure (the ratio between L1 and L2 norms) used in traditional compressed sensing problems, we present DeepHoyer, a set of sparsity-inducing regularizers that are both differentiable almost everywhere and scale-invariant. Our experiments show that enforcing DeepHoyer regularizers can produce even sparser neural network models than previous works, under the same accuracy level. We also show that DeepHoyer can be applied to both element-wise and structural pruning.

To be appeared in ICLR 2020 (poster presentation)

References in corpus (11)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Nearly unbiased variable selection under minimax concave penalty
• Pruning Filters for Efficient ConvNets
• SNIP: Single-shot Network Pruning based on Connection Sensitivity
• A Systematic DNN Weight Pruning Framework using Alternating Direction Method of Multipliers
• Learning Sparse Neural Networks through $L_0$ Regularization
• Euclid in a Taxicab: Sparse Blind Deconvolution with Smoothed l1/l2 Regularization
• Learning Intrinsic Sparse Structures within Long Short-Term Memory
• Soft Filter Pruning for Accelerating Deep Convolutional Neural Networks
• NeST: A Neural Network Synthesis Tool Based on a Grow-and-Prune Paradigm
• Centripetal SGD for Pruning Very Deep Convolutional Networks with Complicated Structure

Cited by in corpus (4)

• PENNI: Pruned Kernel Sharing for Efficient CNN Inference
• SASL: Saliency-Adaptive Sparsity Learning for Neural Network Acceleration
• BWCP: Probabilistic Learning-to-Prune Channels for ConvNets via Batch Whitening
• How Well Do Sparse Imagenet Models Transfer?