Transfer learning

#1

I didn’t get 3rd point

48%20AM%20(2)
48%20AM%20(2)1152×864 141 KB

Also 1st point in
49%20AM
49%20AM1152×864 120 KB

#2

The third point says that when you have very small size dataset, which is very different from the previous one,
then there are two three choices as follows,

  1. We can think of fine tuning our model by adding 3-4 layers freezing initial layers and training last layers, but the original network previously trained contains specific features of previous dataset. So this option is rejected.

  2. We can think of training our model completely without freezing weights of any layers, but since dataset is of very small size we can’t proceed like this as model may overfit.

  3. The only way out is that you can take output of layers somewhere in the starting, as initial layers tries to capture basic features only, so the blog is suggesting to use these outputs to give input to svm classifier, since dataset is quite small and very different.

Hope this helped.

#3

Got it …But
You didn’t explain 1st point in 2nd Screenshot.
what is required for building a chatbot and it’s embedding in website?

#4

Oh yes sorry, my bad
The first point in the second pic says that when we are trying to use transfer learning, we can’t arbitary remove layers from our choice. Secondly it says that, but sometimes we can apply transfer learning on models trained on different image sizes and new dataset images have different size. It would still be fine, since convolutional neural network are capable of capturing all the features and trends. Same holds true to some extent for fully connected layers as well, since fully connected layers takes outputs of that convolutional layers, we can decide to keep some of the fully connected layers as well.

Happy Learning :blush:

#5

ok understood…
what about this?

#6

It’s a very long process, just to have a idea read about it on blogs, like this:

#7

can i make it after finishing this course completely ?

#8

Yes sure, you can built a basic chat bot yourself after completing the course.

#9

ok…

#10

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.