Open satanb4 opened 1 year ago
With testing data, getting accuracy of 99.95% now only on 25 epochs and 0.005 step size:- OpenMP has been implemented. Would be nice to have CUDA accelerated processing, but my CodeSpaces device does not have the CUDA graphics. Maybe I can look into further optimization with parallel processing on CPU as well. Probably with more threads?
Add a training Module for the EMG data
A training module using a neural network would be a great addition for generating custom profiles for users.
This work would benefit the development of EMG based muscle movement classification for users with unique kinds of ailments. PS - Initial Neural Network implementation from scratch done by @satanb4 already in "development" branch