Dear Cyrus
Interesting project. Would you please elaborate more on your basic algorithm and how you are going to optimize it. Moreover, based on your experience, what are the useful libraries in neural network optimization?
Thanks and good luck
A good start would be using MATLAB wizard.
https://www.mathworks.com/products/neural-network.html
Dear Bijan,
One of acceptable manners for optimizing various types of neural networks is a reliable trial and error process. You must run your network over and over till the target is met (i.e., acceptable error). Or you can select the optimum structure by comparing the results obtained from each try. As is clear, the least error shows the best performance but you need to have a assurance that next structures do not perform better. For example, if you draw the RMSE chart, structure D can be selected as the optimum architecture, if the error value is less than structures A, B, C and also structures E, F ... . I think three tries could be sufficient for before and after.
Dear Bijan
Bijan Goshayeshi
Matlab 2018 has provided a great app for classification and regression problems which is user friendly.After using and choosing the best algorithm, as it is case dependent, you can generate the Matlab code and modify it if necessary. I hope it helps.
A picture of app environment is attached.
Please call me if you need anything else +989150464611.
Telegram ID: @CyrusAhmadi
- Badges
- Science topic