-
Notifications
You must be signed in to change notification settings - Fork 5
/
spm_DEM_eval_diff.m
192 lines (168 loc) · 6.34 KB
/
spm_DEM_eval_diff.m
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
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
function [D] = spm_DEM_eval_diff(x,v,qp,M,bilinear)
% evaluates derivatives for DEM schemes
% FORMAT [D] = spm_DEM_eval_diff(x,v,qp,M,bilinear)
% v{i} - casual states
% x(i) - hidden states
% qp - conditional density of parameters
% qp.p{i} - parameter deviates for i-th level
% qp.u(i) - basis set
% qp.x(i) - expansion point ( = prior expectation)
% M - model structure
% bilinear - optional flag to suppress second-order derivatives
%
% D - derivatives
% D.dgdv
% ...
%__________________________________________________________________________
% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
% Karl Friston
% $Id: spm_DEM_eval_diff.m 6734 2016-03-02 12:02:46Z peter $
% check for evaluation of bilinear terms
%--------------------------------------------------------------------------
try
bilinear;
catch
bilinear = 1;
end
% get dimensions
%==========================================================================
nl = size(M,2); % number of levels
ne = sum(spm_vec(M.l)); % number of e (errors)
nx = sum(spm_vec(M.n)); % number of x (hidden states)
np = sum(spm_vec(M.p)); % number of p (parameters)
ny = M(1).l; % number of y (inputs)
nc = M(end).l; % number of c (prior causes)
% initialise cell arrays for hierarchical structure
%--------------------------------------------------------------------------
df.dv = cell(nl - 1,nl - 1);
df.dx = cell(nl - 1,nl - 1);
df.dp = cell(nl - 1,nl - 1);
dg.dv = cell(nl ,nl - 1);
dg.dx = cell(nl ,nl - 1);
dg.dp = cell(nl ,nl - 1);
for i = 1:(nl - 1)
dg.dv{i + 1,i} = sparse(M(i).m,M(i).m);
dg.dx{i + 1,i} = sparse(M(i).m,M(i).n);
dg.dp{i + 1,i} = sparse(M(i).m,M(i).p);
dg.dv{i ,i} = sparse(M(i).l,M(i).m);
dg.dx{i ,i} = sparse(M(i).l,M(i).n);
dg.dp{i ,i} = sparse(M(i).l,M(i).p);
df.dv{i ,i} = sparse(M(i).n,M(i).m);
df.dx{i ,i} = sparse(M(i).n,M(i).n);
df.dp{i ,i} = sparse(M(i).n,M(i).p);
end
if bilinear
for i = 1:(nl - 1)
dg.dvp{i} = cell(M(i).p,1);
dg.dxp{i} = cell(M(i).p,1);
df.dvp{i} = cell(M(i).p,1);
df.dxp{i} = cell(M(i).p,1);
[dg.dvp{i}{:}] = deal(dg.dv);
[dg.dxp{i}{:}] = deal(dg.dx);
[df.dvp{i}{:}] = deal(df.dv);
[df.dxp{i}{:}] = deal(df.dx);
end
end
% Derivatives at each hierarchical level
%==========================================================================
% inline function for evaluating projected parameters
%--------------------------------------------------------------------------
h = @(f,x,v,q,u,p) f(x,v,spm_unvec(spm_vec(p) + u*q,p));
for i = 1:(nl - 1)
% states level i
%----------------------------------------------------------------------
xvp = {x{i},v{i},qp.p{i},qp.u{i},M(i).pE};
% 1st and 2nd partial derivatives (states)
%----------------------------------------------------------------------
if bilinear && np
try
[dgdxp, dgdx] = spm_diff(h,M(i).gx,xvp{:},4,'q');
[dgdvp, dgdv] = spm_diff(h,M(i).gv,xvp{:},4,'q');
[dfdxp, dfdx] = spm_diff(h,M(i).fx,xvp{:},4,'q');
[dfdvp, dfdv] = spm_diff(h,M(i).fv,xvp{:},4,'q');
catch
[dgdxp, dgdx] = spm_diff(h,M(i).g,xvp{:},[2 4],'q');
[dgdvp, dgdv] = spm_diff(h,M(i).g,xvp{:},[3 4],'q');
[dfdxp, dfdx] = spm_diff(h,M(i).f,xvp{:},[2 4],'q');
[dfdvp, dfdv] = spm_diff(h,M(i).f,xvp{:},[3 4],'q');
end
else
try
dgdx = h(M(i).gx,xvp{:});
dgdv = h(M(i).gv,xvp{:});
dfdx = h(M(i).fx,xvp{:});
dfdv = h(M(i).fv,xvp{:});
catch
dgdx = spm_diff(h,M(i).g,xvp{:},2);
dgdv = spm_diff(h,M(i).g,xvp{:},3);
dfdx = spm_diff(h,M(i).f,xvp{:},2);
dfdv = spm_diff(h,M(i).f,xvp{:},3);
end
end
% 1st-order partial derivatives (parameters)
%----------------------------------------------------------------------
try
dfdp = h(M(i).fp,xvp{:});
dgdp = h(M(i).gp,xvp{:});
catch
dfdp = spm_diff(h,M(i).f,xvp{:},4);
dgdp = spm_diff(h,M(i).g,xvp{:},4);
end
% % check which dervatives need to be evaluated
% %====================================================================
% D(i).dgdv = nnz(dgdv) + nnz(spm_vec(dgdvp));
% D(i).dgdx = nnz(dgdx) + nnz(spm_vec(dgdxp));
% D(i).dfdv = nnz(dfdv) + nnz(spm_vec(dfdvp));
% D(i).dfdx = nnz(dfdx) + nnz(spm_vec(dfdxp));
% Constant terms (linking causes over levels)
%----------------------------------------------------------------------
dg.dv{i + 1,i} = -speye(M(i).m,M(i).m);
% place 1st derivatives in array
%----------------------------------------------------------------------
dg.dx{i,i} = dgdx;
dg.dv{i,i} = dgdv;
df.dx{i,i} = dfdx;
df.dv{i,i} = dfdv;
df.dp{i,i} = dfdp;
dg.dp{i,i} = dgdp;
% place 2nd derivatives in array
%----------------------------------------------------------------------
if bilinear && np
for j = 1:length(dgdxp)
dg.dxp{i}{j}{i,i} = dgdxp{j};
dg.dvp{i}{j}{i,i} = dgdvp{j};
df.dxp{i}{j}{i,i} = dfdxp{j};
df.dvp{i}{j}{i,i} = dfdvp{j};
end
end
end
% concatenate hierarchical forms
%==========================================================================
D.dgdv = spm_cat(dg.dv);
D.dgdx = spm_cat(dg.dx);
D.dfdv = spm_cat(df.dv);
D.dfdx = spm_cat(df.dx);
D.dfdp = spm_cat(df.dp);
D.dgdp = spm_cat(dg.dp);
% fixed derivatives w.r.t. prediction errors and states
%--------------------------------------------------------------------------
D.dfdy = sparse(nx,ny);
D.dfdc = sparse(nx,nc);
D.dedy = spm_speye(ne,ny);
D.dedc = -spm_speye(ne,nc,nc - ne);
% bilinear terms if required
%--------------------------------------------------------------------------
if bilinear
D.dgdvp = {};
D.dgdxp = {};
D.dfdvp = {};
D.dfdxp = {};
for i = 1:length(dg.dvp)
for j = 1:length(dg.dvp{i})
D.dgdvp{end + 1} = spm_cat(dg.dvp{i}{j});
D.dgdxp{end + 1} = spm_cat(dg.dxp{i}{j});
D.dfdvp{end + 1} = spm_cat(df.dvp{i}{j});
D.dfdxp{end + 1} = spm_cat(df.dxp{i}{j});
end
end
end