FloodModels/MakeFigure_02_TestMemoryModules.m at main · adredish/FloodModels · GitHub

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
function MakeFigure_02_TestMemoryModules(constructor, varargin)
% MakeFigure_02_TestMemoryModules(constructor)
separate_figures = false;

nU = 500; nP = 10; nB = 10;
process_varargin(varargin);

if ~separate_figures
    figure;
end

R = SimilarityOfRandomPatterns(constructor, nU, 100);

H = constructor(nU);
T = H.myName;

if separate_figures, figure; else subplot(3,1,1); end
S = Test(constructor, nU, nP, 0.1);
imagesc(S); C = colorbar; title(T, 'interpreter', 'none');
caxis([0 1]);
ylabel(C, 'Pattern similarity'); set(C, 'ytick', [0 1]);
xlabel('Patterns'); ylabel('Patterns'); xticks([]); yticks([]);
axis square;
if separate_figures, FigureLayout('layout', [0.5 0.5]); end

if separate_figures, figure; else subplot(3,1,2); end
cla; hold on
switch H.myName
    case 'MM_OuterProduct', N = linspace(0,5,100);
    case 'Hopfield', N = linspace(0,2,100);
    otherwise
        error('Unknown memory module');
end

S = PatternSimilarityNoRecall(constructor, nU, nB, N);
h(1) = ShadedErrorbar(N, nanmean(S), nanstderr(S), 'color','k');
S = PureNoiseTest(constructor, nU, nB, N);
h(2) = ShadedErrorbar(N, nanmean(S), nanstderr(S), 'color','b');
S = PatternNoiseTest(constructor, nU, nB, nP, N);
h(3) = ShadedErrorbar(N, nanmean(S), nanstderr(S), 'color','r');
line(xlim, [0 0], 'color', 'k', 'LineWidth', 2);
legend(h, 'no recall', '1 P stored', sprintf('%d P stored', nP));
yticks([0 1]); ylabel('Pattern similarity');
if separate_figures, FigureLayout('layout', [0.5 0.5]); end

switch H.myName
    case 'MM_OuterProduct',xticks(0:5); xlabel('Noise added (eta * randn())');
    case 'Hopfield', xticks([0 0.5 1 1.5]); xticklabels({'0','0.5','1', 'all'}); xlabel('Noise Added (proportion of bits randomized)');
    otherwise
        error('Unknown memory module');
end

if separate_figures, figure; else subplot(3,1,3); end
cla; hold on; clear h
R = Test_Framing(constructor, 'nU', 500, 'nB', 25);
nP = length(R.p1);
h(1) = errorbar(1:nP, nanmean(R.S(:,:,1)), nanstderr(R.S(:,:,1)), 'b');
h(2) = errorbar(1:nP, nanmean(R.S(:,:,2)), nanstderr(R.S(:,:,2)), 'r');
line(xlim, [0 0], 'color', 'k', 'LineWidth', 2);
legend(h,'Pattern 1', 'Pattern 2');
ylabel('p(recall)');
xlabel('proportion included');
xL = cell(nP,1); for iL = 1:nP; xL{iL} = sprintf('%.2f:%.2f', R.p1(iL)/R.nU, R.p2(iL)/R.nU); end
set(gca, 'XTick', [0, nP/2 nP], ...
    'XTickLabel', arrayfun(@(x,y)sprintf('%.1f:%.1f',x,y), [R.p1(1)/R.nU 0 0], [0 0 R.p2(end)/nU], 'UniformOutput', false));
yticks([0 1]);
if separate_figures, FigureLayout('layout', [0.5 0.5]); end

if ~separate_figures, FigureLayout('layout', [0.5 1]); end

%%

end
%%
function S = Test(constructor, nU, nP, noise)
% constructor, nU, nP, noise
% noise is size 1
H = constructor(nU);
P = H.MakePatterns(nP, nU);
H.AddPatterns(P);
X = nan(size(P)); P0 = nan(size(P));
for iP = 1:size(P,2)
    P0(:,iP) = H.AddNoise(P(:,iP), noise);
    X(:,iP) = H.Recall(P0(:,iP));
end
S = H.PatternSimilarity(X, P);
end

function S = PatternSimilarityNoRecall(constructor, nU, nB, noise)
% constructor nU nBoot noise(list)
nNoise = length(noise);
H = constructor(nU);
S = nan(nB, nNoise);
for iB = 1:nB
    P = H.MakePatterns(1, nU);
    for iN = 1:nNoise
        X = H.AddNoise(P, noise(iN));
        S(iB, iN) = H.PatternSimilarity(P, X);
    end
end
end


function S = PureNoiseTest(constructor, nU, nB, noise)
% constructor nU nBoot noise(list)
nNoise = length(noise);
H = constructor(nU);
S = nan(nB, nNoise);
for iB = 1:nB
    P = H.MakePatterns(1, nU);
    H.AddOnePattern(P);
    for iN = 1:nNoise
        X = H.Recall(H.AddNoise(P, noise(iN)));
        S(iB, iN) = H.PatternSimilarity(P, X);
    end
end
end

function S = PatternNoiseTest(constructor, nU, nB, nP, noise)
% constructor nU nBoot noise(list)
nNoise = length(noise);
H = constructor(nU);
S = nan(nB, nNoise);
for iB = 1:nB
    P = H.MakePatterns(nP, nU);
    H.AddPatterns(P);
    for iN = 1:nNoise
        X = H.Recall(H.AddNoise(P(:,1), noise(iN)));
        S(iB, iN) = H.PatternSimilarity(P(:,1), X);
    end
end
end

function S = SimilarityOfRandomPatterns(constructor, nU, nP)
% constructor nU nB
H = constructor(nU);
P = H.MakePatterns(nP,nU);
S = H.PatternSimilarity(P);
jnoti = tril(true(nP,nP),-1);
S = S(jnoti);
end