I am working in biomedical signal processing. When I am doing classification sensitivity is low. I understand that my data is unbalanced. However It is visible that changing parameter changes the results. So my idea is to choose the best layers and parameters possible. However, my problem lies there I do not know a method which can optimized the Deep ANN. Normally I use matlab. Is there any tool or method to do this?
Performing the preprocessing to your data may be enhance your results.
Generally there is no exact rule to choose the number of layers, activation function ...
A rule of thumb: going deeper (more layers) leads to better results. You have to "play" with these parameters till your reach satisfactory results. The most important stage is data preprocessing as it usually has great implication on the accuracy of your model
what kind of data preprocessing is common in deep ANN? I already used scaling, integration on EEG signal. It doesn't improve. I am using deep belief network. thanks in advance.
- Badges
- Science topic