Paper Info
Title
Fixed Point Optimization Of Deep Convolutional Neural Networks For Object Recognition
Abstract
Deep convolutional neural networks have shown promising results in image and speech recognition applications. The learning capability of the network improves with increasing depth and size of each layer. However this capability comes at the cost of increased computational complexity. Thus reduction in hardware complexity and faster classification arc highly desired. This work proposes an optimization method for fixed point deep convolutional neural networks. The parameters of a pre-trained high precision network are first directly quantized using L2 error minimization. We quantize each layer one by one, while other layers keep computation with high precision, to know the layer-wise sensitivity on word-length reduction. Then the network is retrained with quantized weights. Two examples on object recognition, MNIST and CIFAR-10, are presented. Our results indicate that quantization induces sparsity in the network which reduces the effective number of network parameters and improves generalization. This work reduces the required memory storage by a factor of 1/10 and achieves better classification results than the high precision networks.
Year
Venue
Keywords
2015
2015 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING (ICASSP)
convolutional neural network, quantization, word length optimization, sparsity
Field
DocType
ISSN
MNIST database,Pattern recognition,Convolutional neural network,Computer science,Recurrent neural network,Time delay neural network,Artificial intelligence,Deep learning,Quantization (signal processing),Cognitive neuroscience of visual object recognition,Computational complexity theory
Conference
1520-6149
Citations 
PageRank 
References 
38
3.21
7
Authors
3
Name
Order
Citations
PageRank
Sajid Anwar118419.96
Kyuyeon Hwang223817.67
Wonyong Sung31445166.19