“https://github.com/Jitendra-kaswa/G-2-MLP_basics/blob/master/MLP%20implementation.ipynb”
I implemented MLP as Prateek Bhaiya told, but my model is not so accurate. I didn’t found any error even I tried for different layer sizes, but accuracy didn’t improve.
what can I do to improve efficiency?
hey @Jituk48216 ,
There is no such error in your code. its correct.
Its just that , while we create custom random data using sklearn, we don’t know how the data will be.
Its just at random , so that parameters you choose like noise , etc. to create that data matters a lott.
So to just understand and learn about neural networks , there working just experiment with a number of parameters , like changing number of layer , number of nodes in each layer , learning rate , epochs , etc.
Don’t just go for accuracy ,just understand the concept , if you get that , accuracy will automatically be achieved.
I hope this helps you 
.
sir there is one more thing why we are not taking derivative of softmax function, according to chain rule delta3 should be = (dL/da)*(da/dz),but we are taking only (dL/da) that is (y_-y)
hey @Jituk48216 ,
I guess softmax function is explained in the videos but not yet explained to implement it. Will have a talk about with mentor .
But for now. just for learning purpose you can google about it ,learn and try implementing it on your own.
thank you so much sir 