Introduction
1.Source and Target, Source Domain and Target Domain.
2.Model and Data Distribution.
3.Categories of Transfer Learning.
4.Finance Data
Transfer Learning in Deep Learning Era
1.Main Architecture
2.Independent Feature at Low Layers and Dependent Feature at High Layers
How transferable are features in deep neural networks?
Test the general and specific features in a CNN network
Result of the test
3.A CNN example
Deep Model Based Transfer and Multi-Task Learning forBiological Image Analysis
4.Specific Adaption
(1).user vector -> stock vector
User Vector
(2).partly update the DNN model (input layer, activations of hidden layer, or the output layer)
(3).speaker-dependent layer.
5.Model Transfer
A weak model can be used to teach a stronger model.
Soft target vs Hard target.
