Given brain wave and corresponding sleep stage data set, train a MLP to predict the sleep stage of brain wave sample in test set.
The overall record of this experiment can be found in report/report.pdf
Store the training file and test file to data/train.mat and data/test.mat.
In Matlab®:
>> main;
>> [acc, model, dist] = run_mlp();
>> pred = mlp_predict(model);Run main.m to load data.
Run run_mlp.m to train the MLP. It will return accuracy acc and the predicted
results (including a trained MLP model and the corresponding evaluation metrics
on a 3 by 3 matrix form dist) on cross validation set.
Run mlp_predict.m to predict sleep stage of test samples using the model.
If want to start over:
>> clean
>> main % run main.m script if you want to change training set and cross validation set
>> [acc, model, dist] = run_mlp(); % another round of training
>> pred = mlp_predict(model); % another round of predictionBy default, we use 0.05 of total training set as cross validation set to check the
performance of the model. This can be adjusted in main.m, line 9.
We apply training 10 times on one training set to eliminate the effort of random
initialization of net. This can be adjusted in run_mlp.m, line 26.
The parameters of MLP network can be adjusted in mlp/mlp.m, like network
size, training function, epoch limitation, etc.
Other parameters in PCA, normalization, linear scaling can be further examined
in data_process directory, see details in report/report.pdf.