Mercurial > hg > camir-aes2014
annotate core/tools/machine_learning/cvpartition_trunctrain.m @ 0:e9a9cd732c1e tip
first hg version after svn
| author | wolffd |
|---|---|
| date | Tue, 10 Feb 2015 15:05:51 +0000 |
| parents | |
| children |
| rev | line source |
|---|---|
| wolffd@0 | 1 % --- |
| wolffd@0 | 2 % class cvpartition_trunctrain |
| wolffd@0 | 3 % NOTE: this is a fake cvpartition double for |
| wolffd@0 | 4 % using cvpartitions in truncated-training size experiments |
| wolffd@0 | 5 % --- |
| wolffd@0 | 6 classdef cvpartition_trunctrain |
| wolffd@0 | 7 |
| wolffd@0 | 8 properties (Hidden) |
| wolffd@0 | 9 |
| wolffd@0 | 10 mtest; |
| wolffd@0 | 11 mtraining; |
| wolffd@0 | 12 end |
| wolffd@0 | 13 properties |
| wolffd@0 | 14 N; |
| wolffd@0 | 15 NumTestSets; |
| wolffd@0 | 16 TrainSize; |
| wolffd@0 | 17 TestSize; |
| wolffd@0 | 18 end |
| wolffd@0 | 19 |
| wolffd@0 | 20 |
| wolffd@0 | 21 methods |
| wolffd@0 | 22 |
| wolffd@0 | 23 % --- |
| wolffd@0 | 24 % constuctor: directly calculates the truncated testset |
| wolffd@0 | 25 % --- |
| wolffd@0 | 26 function P = cvpartition_trunctrain(Pin, perctrain) |
| wolffd@0 | 27 |
| wolffd@0 | 28 P.N = Pin.N; |
| wolffd@0 | 29 P.NumTestSets = Pin.NumTestSets; |
| wolffd@0 | 30 |
| wolffd@0 | 31 for i = 1:Pin.NumTestSets |
| wolffd@0 | 32 |
| wolffd@0 | 33 % copy testing data |
| wolffd@0 | 34 P.TestSize(i) = Pin.TestSize(i); |
| wolffd@0 | 35 P.mtest{i} = Pin.test(i); |
| wolffd@0 | 36 |
| wolffd@0 | 37 % calculate new training size |
| wolffd@0 | 38 P.TrainSize(i) = ceil(perctrain * Pin.TrainSize(i)); |
| wolffd@0 | 39 |
| wolffd@0 | 40 % get actual training indices |
| wolffd@0 | 41 idx = find(Pin.training(i)); |
| wolffd@0 | 42 |
| wolffd@0 | 43 % --- |
| wolffd@0 | 44 % TODO: save the permutation in a global variable, |
| wolffd@0 | 45 % tomake the same smaller set available |
| wolffd@0 | 46 % for all further experiments. |
| wolffd@0 | 47 % moreover, it would be great if the smaller training sets |
| wolffd@0 | 48 % are subsets of the bigger ones |
| wolffd@0 | 49 % --- |
| wolffd@0 | 50 tokeep = randperm(numel(idx)); |
| wolffd@0 | 51 tokeep = tokeep(1:P.TrainSize(i)); |
| wolffd@0 | 52 |
| wolffd@0 | 53 % get indices to keep |
| wolffd@0 | 54 idx = idx(tokeep); |
| wolffd@0 | 55 |
| wolffd@0 | 56 % build truncated training set |
| wolffd@0 | 57 P.mtraining{i} = false(P.N, 1); |
| wolffd@0 | 58 P.mtraining{i}(idx) = true; |
| wolffd@0 | 59 end |
| wolffd@0 | 60 end |
| wolffd@0 | 61 |
| wolffd@0 | 62 function out = test(P, i) |
| wolffd@0 | 63 |
| wolffd@0 | 64 out = P.mtest{i}; |
| wolffd@0 | 65 end |
| wolffd@0 | 66 |
| wolffd@0 | 67 function out = training(P, i) |
| wolffd@0 | 68 |
| wolffd@0 | 69 out = P.mtraining{i}; |
| wolffd@0 | 70 end |
| wolffd@0 | 71 |
| wolffd@0 | 72 end |
| wolffd@0 | 73 end |