brain_navigation/DT_margin_test.m at master · epbennetts/brain_navigation · 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
close all force;
clear all;

set(0,'DefaultAxesColorOrder','default');
%target = rand(1000,1)+1;
%nonTarget = rand(100000,1);

%make dummy data 1
targetCentre = 2;
nonTargetCentre = -2;
targetSize = 1000;
nonTargetSize = 10000;
target1 = normrnd(targetCentre,0.9,targetSize, 1);
nonTarget1 = normrnd(nonTargetCentre,0.9,nonTargetSize, 1);
allData1 = [target1; nonTarget1];
% %make it look balanced
% modifier = 5;
% x_repetitions = nonTargetSize/targetSize - modifier; %assumes target is smallest
% y_repetitions = 1;
% target_magnified = repmat(target1, x_repetitions, y_repetitions);
% allData1 = [target_magnified; nonTarget1];


%make dummy data 2
sep = 2;
nonTarget2 = normrnd(nonTargetCentre-sep,1.2,nonTargetSize, 1);
target2 =  normrnd(targetCentre+sep,1.3,targetSize, 1);
allData2 = [target2; nonTarget2];

%set classes
classes = zeros(size(target1,1)+size(nonTarget1,1),1);
classes(1:size(target1,1),:) = 1;

%set cost function (balanced)
costFunc = ComputeBalancedCostFunc(classes);

%train
tree1 = fitctree(allData1, classes, 'MaxNumSplits', 1, 'cost', costFunc, 'ClassNames', [1,0]);
tree2 = fitctree(allData2, classes, 'MaxNumSplits', 1, 'cost', costFunc, 'ClassNames', [1,0]);

% Store thresholds
%unbalanced
threshold_raw = tree1.CutPoint; %threshold with NaNs
threshold1 = threshold_raw(~isnan(threshold_raw));
%balanced
threshold_raw = tree2.CutPoint; %threshold with NaNs
threshold2 = threshold_raw(~isnan(threshold_raw));

%test
predictedLabels = predict(tree2, allData2);

%calc confusion mat & bal accuracy
[confMat,balAcc] = ComputeConfusion(classes,predictedLabels);


%view
% view(tree1,'Mode','graph')
% view(tree2,'Mode','graph')

%PLOT 1
%gene expr histograms
%UNBALANCED
figure()
hold on;
box on;
%title('Gene expression');
histogram(target1, 'Normalization', 'count', 'BinWidth', 0.2, 'Facecolor', [0.2 0.8 0.2]);
histogram(nonTarget1, 'Normalization', 'count', 'BinWidth', 0.1, 'FaceColor', [0.9 0.3 0.3]);
set(gca,'fontsize', 12);
%xticks(-5:1:6)
%thresholds
x1 = xline(threshold1, '--r', 'LineWidth', 1.5);
%xline(threshold_bal, '--g', 'LineWidth', 1);
%legend
legend({'target','non-target','threshold'}, 'Location','northeast');
title('Small margin')
xlabel('gene expression')
ylabel('counts')
xlim([-8,8])
ylim([0,500])
hold off;

%PLOT 2
%BALANCED
%-- histograms but with target modified to look like it has similar counts
%(Except it looks as if there are 2x as many targets because we're repeating
%the exact same examples...)
%gene expr histograms
figure()
hold on;
box on;
%title('Gene expression');
histogram(target2, 'Normalization', 'count', 'BinWidth', 0.2, 'Facecolor', [0.2 0.8 0.2]);
histogram(nonTarget2, 'Normalization', 'count', 'BinWidth', 0.1, 'FaceColor', [0.9 0.3 0.3]);
set(gca,'fontsize', 12);
%xticks(-5:1:6)
%thresholds
%xline(threshold_unbal, '--r', 'LineWidth', 1);
xline(threshold2, '--g', 'LineWidth', 1.5);
%legend
legend({'target','non-target','threshold'}, 'Location','northeast');
title('Large margin')
xlabel('gene expression')
ylabel('counts')
xlim([-8,8])
ylim([0,500])
hold off;