Pyramid Vector Quantization and Bit Level Sparsity in Weights for Efficient Neural Networks Inference
This work addresses computational efficiency for CNN deployment, though it appears incremental as it builds on existing quantization methods.
The paper tackles efficient neural network inference by introducing Pyramid Vector Quantization (PVQ) to create sparse and compressible CNN weights, enabling multiplier elimination while maintaining high performance, as demonstrated on Tiny Yolo v3.
This paper discusses three basic blocks for the inference of convolutional neural networks (CNNs). Pyramid Vector Quantization (PVQ) is discussed as an effective quantizer for CNNs weights resulting in highly sparse and compressible networks. Properties of PVQ are exploited for the elimination of multipliers during inference while maintaining high performance. The result is then extended to any other quantized weights. The Tiny Yolo v3 CNN is used to compare such basic blocks.