NeuralNetProject/layer.m at master · Ferruolo/NeuralNetProject · 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
classdef layer
    %LAYER Summary of this class goes here
    %   Detailed explanation goes here

    properties
        name
        activation_fxn
        weights
        bias
        grad
        output
    end

    methods
        function obj = layer(name, activation_fxn, input, output, bias_shape)
            obj.name = name;
            obj.weights = tensor(rand(input, output));
            obj.bias = tensor(rand(bias_shape));
            obj.activation_fxn = activation_fxn;
        end

        function x = lay_forward(obj, x)
             disp("_-_-_-_-_-_-_-_-_-_-_-_-_-_");

            x = mult(x, obj.weights, obj.name);
%             x = k_mult(x, obj.weights, obj.name);
% %             x = sum_2d(x, obj.name);
%              disp("add");
%              disp(x.value);
%              disp("bias");
%              disp(obj.bias.value);

            x = k_add(x, obj.bias, obj.name);
%              %activation_fxn must be manual update
%              disp("logistic");
%              disp(x.value);

            z = obj.activation_fxn.fxn_forward(x.value);
            x = update(x, z);
%              disp("out");
%              disp(x.value);

            x.graph(length(x.graph) + 1) = graph_node(obj.activation_fxn.name,  x, obj.weights, obj.name, false);

        end
        function obj = lay_backward_mul(obj, update_fxn, running_grad)
%             disp("shape");
%             shape = size(obj.weights.value);
%             disp(shape);
%             disp("run")
%             disp(size(running_grad));

            obj.weights = bprop_compute_tensor(update_fxn, running_grad, obj.weights);


%             for i = 1:shape(3)
%                 run = running_grad(:, :, i);
%                 w = obj.weights.value(:, :, i);
%
%                 obj.weights.value(:, :, i) = bprop_compute_matrix(update_fxn, run, w );
%             end
%
%             obj.weights = bprop_compute(update_fxn, running_grad, obj.weights);
%              obj.bias = bprop_compute(update_fxn, running_grad, obj.bias);


        end
        function obj = lay_backward_add(obj, update_fxn, running_grad)
%             obj.weights = bprop_compute(update_fxn, running_grad, obj.weights);
            obj.bias = bprop_compute_tensor(update_fxn, running_grad, obj.bias);


        end

    end
end