calculate training coeffecient of determination r^2 and mean absolute error

Hi everyone,I hope this message finds you well. I am currently working on a prediction task using LSTM, and I have successfully obtained the training RMSE using info in following codde

[net info] = trainNetwork(xtrain, ytrain, layers, options);

i have also obtained testing metrics using their formulas

   
    Y = predict(net, xtest);
    e = (ytest - Y);
    rmse = mean(sqrt(mean((Y - ytest).^2)));
    mae1 = mae(e);
    Rsq1 = 1 - sum((ytest - Y).^2) / sum((ytest - mean(ytest)).^2);
    mse = mean(mean((ytest - Y).^2));

However, I am curious to know if there is a straightforward way to retrieve additional training metrics such as R² and MAE for the LSTM model. Your insights and guidance on this matter would be greatly appreciated.

Thank you in advance for your time and assistance.

Answer 1

[Debraj Maji]

Hi @Mahi,

I understand that you are trying to retrieve additional training metrics for the aforementioned LSTM Model. As of 2023b the available metrics for tracking are:

• AccuracyMetric
• AUCMetric
• FScoreMetric
• PrecisionMetric
• RecallMetric
• RMSEMetric

One of the ways to track R-squared and MAE during training is by creating a custom Deep Learning Metric Object and specifying it in 'trainingOptions' under Metrics argument. The steps to create a Deep Learning Metric Object can be found here: https://in.mathworks.com/help/deeplearning/ug/define-custom-deep-learning-metric.html

For more info on options available during training you can refer to the following documentation:

https://in.mathworks.com/help/deeplearning/ref/trainingoptions.html

I hope this resolves your query.

With regards,

Debraj.

[Mahi]

Right . Thank you very much