BMVC/eval_model.m at master · csrhddlam/BMVC · 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
label_eval = 23;
probes = cell(3, 1);

probes{1} = find((partition_all == 1) .* (label_all == label_eval));

for i = 1:length(label_val)
    probes{ i + 1 } = find((partition_all == 2) .* (label_all == label_val(i)));
end

results_raw = cell(size(probes));
results_mean = cell(size(probes));

results_min = 99999999;
results_max = -99999999;
for probe_index = 1:length(probes) % for each probe
    disp(probe_index);
    eval_indices = probes{probe_index};
    results_raw{probe_index} = zeros(length(eval_indices), const_eval_samples);
    for eval_index_index = 1:length(eval_indices) % for each data point
        eval_index = eval_indices(eval_index_index);
        for sample_index = 1:const_eval_samples
            visible_units = get_data_from_index(data_all, eval_index, const_h, const_w, channels);
            hidden_units = double(randn(hidden, 1) < 0.5);
            probability = [visible_units; hidden_units]' * matrix(1:visible+hidden,:) * [visible_units; hidden_units;1];
            results_raw{probe_index}(eval_index_index, sample_index) = gather(probability);
        end
    end
    results_mean{probe_index} = mean(results_raw{probe_index}, 2);
    results_min = min(results_min, min(results_mean{probe_index}));
    results_max = max(results_max, max(results_mean{probe_index}));
end

hist_bins = results_min:((results_max - results_min) / 100):results_max;
zero_bins = [-200, 1];
results_hist = zeros(length(probes),length(hist_bins));
results_zero = zeros(length(probes),length(zero_bins));
for probe_index = 1:length(probes)
    results_hist(probe_index,:) = hist(results_mean{probe_index}, hist_bins);
    results_zero(probe_index,:) = hist(results_mean{probe_index}, zero_bins);
end
figure;
plot(hist_bins, results_hist);
figure;
plot(zero_bins, results_zero);

% FN = results_zero(23, 1);
% TP = results_zero(23, 2);
% temp = sum(results_zero, 1) - results_zero(23, :);
% TN = temp(1);
% FP = temp(2);
% precision = TP / (TP + FP)
% recall = TP / (TP + FN)
% accuracy = (TP + TN) / (TP + TN + FP + FN)
% norm_accuracy = TP / (TP + FN) / 2 + TN / (TN + FP) / 2