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
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
|
function hdv = mlphdotv_weighted(net, x, t, eso_w, v)
%MLPHDOTV Evaluate the product of the data Hessian with a vector.
%
% Description
%
% HDV = MLPHDOTV(NET, X, T, V) takes an MLP network data structure NET,
% together with the matrix X of input vectors, the matrix T of target
% vectors and an arbitrary row vector V whose length equals the number
% of parameters in the network, and returns the product of the data-
% dependent contribution to the Hessian matrix with V. The
% implementation is based on the R-propagation algorithm of
% Pearlmutter.
%
% See also
% MLP, MLPHESS, HESSCHEK
%
% Copyright (c) Ian T Nabney (1996-9)
% Check arguments for consistency
errstring = consist(net, 'mlp', x, t);
if ~isempty(errstring);
error(errstring);
end
ndata = size(x, 1);
[y, z] = mlpfwd(net, x); % Standard forward propagation.
zprime = (1 - z.*z); % Hidden unit first derivatives.
zpprime = -2.0*z.*zprime; % Hidden unit second derivatives.
vnet = mlpunpak(net, v); % Unpack the v vector.
% Do the R-forward propagation.
ra1 = x*vnet.w1 + ones(ndata, 1)*vnet.b1;
rz = zprime.*ra1;
ra2 = rz*net.w2 + z*vnet.w2 + ones(ndata, 1)*vnet.b2;
switch net.actfn
case 'softmax' % Softmax outputs
nout = size(t, 2);
ry = y.*ra2 - y.*(sum(y.*ra2, 2)*ones(1, nout));
otherwise
error(['Unknown activation function ', net.actfn]);
end
% Evaluate a weighted delta for the output units.
temp = y - t;
for m=1:ndata,
delout(m,:)=eso_w(m,1)*temp(m,:);
end
clear temp;
% Do the standard backpropagation.
delhid = zprime.*(delout*net.w2');
% Now do the R-backpropagation.
rdelhid = zpprime.*ra1.*(delout*net.w2') + zprime.*(delout*vnet.w2') + ...
zprime.*(ry*net.w2');
% Finally, evaluate the components of hdv and then merge into long vector.
hw1 = x'*rdelhid;
hb1 = sum(rdelhid, 1);
hw2 = z'*ry + rz'*delout;
hb2 = sum(ry, 1);
hdv = [hw1(:)', hb1, hw2(:)', hb2];
|
