Reducing computations in a neural network

Patent metadata

Field	Value
Publication number	US-2016358068-A1
Application number	US-201615014686-A
Country	US
Kind code	A1
Filing date	Feb 3, 2016
Priority date	Jun 4, 2015
Publication date	Dec 8, 2016
Grant date	—

Abstract

Reducing computations in a neural network may include determining a group including a plurality of convolution kernels of a convolution stage of a neural network. The convolution kernels of the group are similar to one another. A base convolution kernel for the group may be determined. Scaling factors for a plurality of input feature maps processed by the group may be calculated. The convolution stage of the neural network may be modified to calculate a composite input feature map using the scaling factors and apply the base convolution kernel to the composite input feature map.

First claim

What is claimed is:

1 . A method, comprising:
determining, using a processor, a group comprising a plurality of convolution kernels of a convolution stage of a neural network, wherein the convolution kernels of the group are similar to one another; determining a base convolution kernel for the group; determining scaling factors for a plurality of input feature maps processed by the group; and modifying the convolution stage to calculate a composite input feature map using the scaling factors and apply the base convolution kernel to the composite input feature map.

2 . The method of claim 1 , wherein determining the group of convolution kernels comprises:
determining a similarity metric between a first convolution kernel and a second convolution kernel; and including the first convolution kernel and the second convolution kernel in the group responsive to determining that the similarity metric meets a similarity criterion.

3 . The method of claim 2 , wherein determining a similarity metric comprises:
calculating a ratio of the first convolution kernel to the second convolution kernel; and calculating a standard deviation for the ratio of the first convolution kernel and the second convolution kernel.

4 . The method of claim 1 , wherein determining the group of convolution kernels comprises:
selecting a convolution kernel from among a plurality of convolution kernels that operate on the feature map as a primary convolution kernel; and calculating a ratio of the primary convolution kernel to other convolution kernels of the plurality of convolution kernels not yet assigned to a group.

5 . The method of claim 1 , wherein determining the base convolution kernel for the group comprises:
selecting a convolution kernel of the group as the base convolution kernel according to a base selection metric.

6 . The method of claim 1 , wherein determining the base convolution kernel for the group comprises:
determining the base convolution kernel as a function of the convolution kernels of the group.

7 . The method of claim 1 , further comprising:
retraining the neural network comprising the modified convolution stage without permitting changes to membership in the group of convolution kernels.

8 . The method of claim 1 , further comprising:
executing the neural network comprising the modified convolution stage, wherein executing the neural network comprises: scaling each of the plurality of input feature maps using the scaling factors, generating the composite input feature map as a sum of the scaled input feature maps, and applying the base convolution kernel to the composite input feature map.

9 . A system, comprising:
a processor programmed to initiate executable operations comprising: determining a group comprising a plurality of convolution kernels of a convolution stage of a neural network, wherein the convolution kernels of the group are similar to one another; determining a base convolution kernel for the group; determining scaling factors for a plurality of input feature maps processed by the group; and modifying the convolution stage to calculate a composite input feature map using the scaling factors and apply the base convolution kernel to the composite input feature map.

10 . The system of claim 9 , wherein determining the group of convolution kernels comprises:
determining a similarity metric between a first convolution kernel and a second convolution kernel; and including the first convolution kernel and the second convolution kernel in the group responsive to determining that the similarity metric meets a similarity criterion.

11 . The system of claim 10 , wherein determining a similarity metric comprises:
calculating a ratio of the first convolution kernel to the second convolution kernel; and calculating a standard deviation for the ratio of the first convolution kernel and the second convolution kernel.

12 . The system of claim 9 , wherein determining the group of convolution kernels comprises:
selecting a convolution kernel from among a plurality of convolution kernels that operate on the feature map as a primary convolution kernel; and calculating a ratio of the primary convolution kernel to other convolution kernels of the plurality of convolution kernels not yet assigned to a group.

13 . The system of claim 9 , wherein determining the base convolution kernel for the group comprises:
selecting a convolution kernel of the group as the base convolution kernel according to a base selection metric.

14 . The system of claim 9 , wherein determining the base convolution kernel for the group comprises:
determining the base convolution kernel as a function of the convolution kernels of the group.

15 . The system of claim 9 , wherein the processor is further programmed to initiate executable operations comprising:
retraining the neural network comprising the modified convolution stage without permitting changes to membership in the group of convolution kernels.

16 . A non-transitory computer-readable storage medium having instructions stored thereon which, when executed by a processor, perform a method comprising:
determining, using the processor, a group comprising a plurality of convolution kernels of a convolution stage of a trained neural network, wherein the convolution kernels of the group are similar to one another; determining a base convolution kernel for the group; determining scaling factors for a plurality of input feature maps processed by the group; and modifying the convolution stage to calculate a composite input feature map using the scaling factors and apply the base convolution kernel to the composite input feature map.

17 . The non-transitory computer-readable storage medium of claim 16 , wherein determining the group of convolution kernels comprises:
determining a similarity metric between a first convolution kernel and a second convolution kernel; and including the first convolution kernel and the second convolution kernel in the group responsive to determining that the similarity metric meets a similarity criterion.

18 . The non-transitory computer-readable storage medium of claim 17 , wherein determining a similarity metric comprises:
calculating a ratio of the first convolution kernel to the second convolution kernel; and calculating a standard deviation for the ratio of the first convolution kernel and the second convolution kernel.

19 . The non-transitory computer-readable storage medium of claim 16 , wherein determining the group of convolution kernels comprises:
selecting a convolution kernel from among a plurality of convolution kernels that operate on the feature map as a primary convolution kernel; and calculating a ratio of the primary convolution kernel to other convolution kernels of the plurality of convolution kernels not yet assigned to a group.

20 . The non-transitory computer-readable storage medium of claim 16 , wherein the method further comprises:
retraining the neural network comprising the modified convolution stage without permitting changes to membership in the group of convolution kernels.

Assignees

• Samsung Electronics Co Ltd

Inventors

• Brothers John W
• Lee Joohoon

Classifications

• G06N3/045 Primary

  Combinations of networks · CPC title

• G06N3/082

  modifying the architecture, e.g. adding, deleting or silencing nodes or connections · CPC title

• G06V10/82 Primary

  using neural networks · CPC title

• G06N3/0464 Primary

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N99/005

  Physics · mapped topic