Stacked Filters Stationary Flow For Hardware-Oriented Acceleration Of Deep Convolutional Neural Networks

01/23/2018

∙

To address memory and computation resource limitations for hardware-oriented acceleration of deep convolutional neural networks (CNNs), in this paper we present a computation flow, stacked filters stationary flow (SFS), and a corresponding data encoding format, relative indexed compressed sparse filter format (CSF), and also a three dimensional Single Instruction Multiple Data (3D-SIMD) processor architecture to take full advantage of these two features. Comparing with the state-of-the-art result (Han et al., 2016b), our method achieved 1.11x improvement in reducing the storage required by AlexNet, and 1.09x improvement in reducing the storage required by SqueezeNet, without loss of accuracy on the ImageNet dataset. Moreover, using this approach, chip area for logics handling irregular sparse data access can be saved.

READ FULL TEXT

Stacked Filters Stationary Flow For Hardware-Oriented Acceleration Of Deep Convolutional Neural Networks

FPGA Implementations of 3D-SIMD Processor Architecture for Deep Neural Networks Using Relative Indexed Compressed Sparse Filter Encoding Format and Stacked Filters Stationary F

Distilling Critical Paths in Convolutional Neural Networks

Sparq: A Custom RISC-V Vector Processor for Efficient Sub-Byte Quantized Inference

Jointly Sparse Convolutional Neural Networks in Dual Spatial-Winograd Domains

CodeX: Bit-Flexible Encoding for Streaming-based FPGA Acceleration of DNNs

Intra-layer Nonuniform Quantization for Deep Convolutional Neural Network

Learning on the Edge: Explicit Boundary Handling in CNNs

Stacked Filters Stationary Flow For Hardware-Oriented Acceleration Of Deep Convolutional Neural Networks

Related Research

FPGA Implementations of 3D-SIMD Processor Architecture for Deep Neural Networks Using Relative Indexed Compressed Sparse Filter Encoding Format and Stacked Filters Stationary F

Distilling Critical Paths in Convolutional Neural Networks

Sparq: A Custom RISC-V Vector Processor for Efficient Sub-Byte Quantized Inference

Jointly Sparse Convolutional Neural Networks in Dual Spatial-Winograd Domains

CodeX: Bit-Flexible Encoding for Streaming-based FPGA Acceleration of DNNs

Intra-layer Nonuniform Quantization for Deep Convolutional Neural Network

Learning on the Edge: Explicit Boundary Handling in CNNs