Does Transfer Learning work well for other machine learning models? If possible, could you please provide some literature references? Thank you very much for your help!
Transfer learning works in the cases where the base features for the given input are more or less same. Taking the example of Dogs vs Cats classification, we take models like Resnet or VGG because the earlier layers of these networks extract features like edges etc. As we move towards the higher layers, the features extracted become more abstract like ear or nose or faces etc.
So transfer learning would work in any domain where the base features are the same. One example would be to perform activity recognition using sensors data such as accelerometer or gyroscope. if you have a neural net to perform activity recognition based on sensor data, you can finetune the last few layers to use the model to classify the gestures into different categories.
Thanks a lot!