Implementation of Stanfords ConvNetJS to calculate relations between EMG signals of a handsensor and the applied force
This network allows to import data as a csv-seperated file.
The trained network may be exported as JSON using the "Save Network" button.
In this example, the applied force gets calculated based on the measured EMG-signals.
For debugging purposes, the actual force output, the force output calculated by the neural network as well as the training accuracy of the current iteration are displayed. To gain an overview over the networks performance, the amount of total iterations and the accuracy during the last X iterations are displayed.
To visualize the development of the neural network, additional statistics are available as graphs.
Loss-graph: Respresents the loss between every batch of iterations.
Accuracy-graph: Represents the development of the accuracy with increasing number of iterations. This allows to determine whether the performance is converging or further training may improve the accuracy of the neural network.