Posts Activity Stream
Usually that is the difference between the expected and calculated values. Sometimes it is used as a scaling factor (such as learning rate) for the weight matrix update.
Should there be an end difference for the correlation co-efficient based on the activation function? I can get 0.94 for tanh, but only 0.64 for logistic.
When the testing and training data are different and the respective input files are not standardized, in that case is this the right way of preprocessing:
preprocess non_standardized_training_input_file.txt non_standardized_training_input_file.txt 1 standardized_training_input_file.txt
preprocess non_standardized_training_input_file.txt non_standardized_testing_input_file.txt 1 standardized_testing_input_file.txt
How is training error calculated while training the network?
Preprocessing testing and training input file
in Data Preprocessing
agreed. what's the question?