1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
|
function Y = dnn_builtin (W, bias, Y0)
%DNN_BUILTIN Sparse deep neural network without @GrB
% Performs ReLU inference using input feature vector(s) Y0, DNN weights W,
% and bias vectors.
%
% Slightly revised from graphchallenge.org.
%
% Usage:
%
% Y = dnn_builtin (W, bias, Y0)
%
% See also GrB.dnn, dnn_builtin2gb.
% note: this is now ported to Octave, by avoiding the use of singleton
% expansion.
% SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
% SPDX-License-Identifier: Apache-2.0
Y = Y0 ;
n = size (Y, 2) ;
for i=1:length(W)
% Propagate through layer.
Z = Y * W {i} ;
% Apply bias to non-zero entries.
b = spdiags (bias{i}', 0, n, n) ;
Y = Z + (double(logical(Z)) * b) ;
% Threshold negative values.
Y (Y < 0) = 0 ;
% Threshold maximum values.
Y (Y > 32) = 32 ;
end
|
