Hi Shailaja,
The function 'learngd()' implements gradient descent as a learning algorithm, primarily used for training shallow neural networks in MATLAB’s Neural Network Toolbox.
However, l'earngd()' is not well-suited for deep learning architectures, because It is a basic gradient descent function that does not include momentum or adaptive learning rate adjustments, which are crucial for training deep networks effectively. For deep learning models, you can use more advanced optimizers like Adam, RMSprop, or L-BFGS to efficiently update weights in deeper architectures. Some other functions similar to ‘learngd’ are ‘learngdm’,’learnp’, ‘learnk’, etc.
For deep learning, use newer training options like 'adam', 'sgdm', or 'rmsprop' via the 'trainingOptions' function. For more information, refer to the following documentations:
- learngdm: https://www.mathworks.com/help/deeplearning/ref/learngdm.html
- learnp: https://www.mathworks.com/help/deeplearning/ref/learnp.html
- learnk: https://www.mathworks.com/help/deeplearning/ref/learnk.html
- trainingOptions: https://www.mathworks.com/help/deeplearning/ref/trainingoptions.html
Hope this helps!
