Doubt in lecture

#1

at 2:30 in the video, prateek bhaiya increased the number of filters in the 3rd and the 4th convolution layer, ie from 64 to 128 filters, although it however increased the number of parameters , shouldn’t we try to reduce the number of parameters in the model?

#2

Firstly you need to understand for what these filters are actually used for.

Convolutional layers are used to extract features from the images and as we go deeper in a model we need to find more features. These filters have the main role to exract these features.

For Example , in first layer we were able to find features like edges ,corners ,etc. In the second layer we were able to find features like shapes,etc.

So as we move deeper in a convolutional network we are extracting more complex features and hence we require more number of filters to perform that task and hence we need to increase those number of filters in subsequent layers.

#3

ok sir i got that! but i guess once prateek bhaiya said that the number of parameters are quite high in comparison to the data available in the file, so we should try to reduce the number of parameters, So my question is that while creating a sequential model, should we try to reduce the number of parameters , or it doesn’t matter ie the number of parameters is not a factor while creating a sequential model?

#4

hey @saksham_thukral,
The answer to your question is :
Yes number of parameters does matter in a deep learning model. You should try to reduce them only till your results are good enough to be used .

Theses points might help you understand it better:

  1. Very Less no. of Parameters : May be lead to underfitting.
  2. Somewhat accurate no.of parameters : Model is correctly fitted.
  3. Very large no. of parameters : May lead to overfitting.
#5

Hey @saksham_thukral ,
i guess your doubt about number of parameters is now resolved , if it so , then kindly mark this doubt as resolved in the panel and also provide your valuable feedback as it will help to us improve our working system to work more effectively and clear doubts much easily.

Thank You,
Prashant Arora

#6

I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.

On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.