RAAM/Module.m at master · gregwayne/RAAM · 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
classdef Module

    properties

        enc;
        dec;

        NH;
        NC;
        NL;

        init_flag;
        code_init;
        eta_init;

        high;
        code;
        low;

        mse;

    end

    methods


        function obj = Module(NL,NC,NH,NHid,g,eta_e,eta_d,eta_init)

            N0          = 1 + NL + NC + NH;
            N1          = NH;
            N2          = NC;

            obj.enc         = SNet(N0,NHid,N2,g,eta_e);
            obj.dec         = SNet(1+NC+NH,NHid,NL+NC,g,eta_d);

            obj.NH          = NH;
            obj.NC          = NC;
            obj.NL          = NL;
            obj.eta_init    = eta_init;

            obj.code_init   = zeros(NC,1);
            obj.low         = zeros(NL,1);

            obj.mse = 0;

        end

        function obj = SetInputs(obj,low,high)

            obj.low     = low;
            obj.high    = high;

        end

        function obj    = PropagateLearn(obj,niter,learn_flag,eta_rec)

            code_o      = LinearThreshold(obj.code);
            xIn         = [1;obj.low;code_o;obj.high];
            obj.enc     = obj.enc.FProp(xIn);
            code        = LinearThreshold(obj.enc.X2);

            if ~learn_flag
                niter = 1;
            end

            target          = LinearThreshold([obj.low;code_o]);

            p               = {};
            p.X0            = [1;code;obj.high];
            p.eta_min       = eta_rec;
            p.iters         = niter;
            p.constraint    = p.X0;
            p.cw            = 1;
            p.target        = target;
            p.cidxs         = 2:(1+obj.NC);

            obj.dec         = obj.dec.Minimize(p);
            obj.mse         = norm(target-obj.dec.X2);

            if learn_flag

                obj.dec         = obj.dec.ErrorLearn(target-obj.dec.X2,1);
                obj.dec         = obj.dec.Tag();
                obj.dec         = obj.dec.Untag();

                obj.enc         = obj.enc.FProp(xIn);
                ch              = LinearThreshold(obj.dec.X0(2:1+obj.NC));
                obj.enc         = obj.enc.ErrorLearn(ch-obj.enc.X2,1);
                obj.enc         = obj.enc.Tag();
                obj.enc         = obj.enc.Untag();

                if obj.init_flag
                    obj.code_init   = obj.code_init + obj.eta_init*(obj.dec.X2(1+obj.NL:end)-obj.code_init);
                    obj.init_flag   = 0;
                end

            end

            obj.code    = obj.enc.X2;

        end

        function code = Encode(obj)

            code                = obj.enc.FastProp([1;obj.low;obj.code]);

        end

        function [low,code_o] = Decode(obj,code,high)

            out                 = LinearThreshold(obj.dec.FastProp([1;code;high]));
            low                 = out(1:obj.NL);
            code_o              = out(obj.NL+1:end);

        end

        function obj = ResetCode(obj)

            obj.code            = obj.code_init;
            obj.low             = 0*obj.low;
            obj.high            = 0*obj.high;
            obj.init_flag       = 1;

        end

    end

end

function y = LinearThreshold(x)

    y = ((x > -1).*(x < 1)).*x - (x < -1) + (x > 1);

end