as we have seen in logistic/linear and many more ML algorithm we know that we have to learn theta and we use Gradient descent algorithm to learn them .In MLP we learn weights and bias by gradient descent .but what exactly are we learning in CNN and by which algorithm are we learning these parameter(ex - gradient descent for linear regression).
What exactly does a CNN model learns
hey @Par1hsharma ,
so to extract information from images , we need to use filters like gaussian filters and all.
But in Deep learning , we can’t like specify this as a filter.
So , we give a number of filters ( default filters ) , that are applied on the images to learn that information. and then with backpropagation they update there weights of learning those information values.
These things in CNN are very deep , and won’t be possible to explain here.
So you can search them on google , and then let me know if you have any doubts.
can you please share a good YouTube video where I can visually see what actually a CNN is doing because in these tutorial we are just adding layers and writing code to compile and fit . but what is the algorithm which learns filters and how does backpropogragion works in CNN?
What exactly are they learning . Is there a good video where I can get this information and visually see it ?
You can try these
I hope they help.
These things are very deep in nature and this course is made to just give you an understanding about the topics and all. No issues , you can try and understand it , if still there is some doubt then let me know.