Inception network motivation

We have learnt till now that number of channels remain unchanged after applying pooling.
But in the above lecture the number of channels was reduced from 192 to 32. Why after applying pooling the number of channels has changed.
lecture link - (

I think you post it under the wrong course. However, I would like to know the answer too. :grinning:

Hello @ccc888 and @atulpandey,
Thank you for your question.

This is an interesting question, many learners have asked this question before.
Actually, Andrew meant that the max-pooling is applied to the previous convolutional layer (the one that has a 5x5 filter). You can see that because he counted the total number of layers just after adding the max-pooling layer.

This is not a usual practice for inception model, usually, the convolution extracts the outcome of the Depth concatenation layer or the max-pooling layer. This was just an example of what is possible to do with data. Later in the same lecture

I would love to answer any more questions you might have.