Visualizing Transfer Learning
This work provides insights into the transfer learning process for researchers in computer vision and deep learning, but it is incremental as it focuses on visualization rather than new methods or applications.
The researchers visualized individual neurons in a deep image recognition network during transfer learning, qualitatively demonstrating novel properties such as adaptation speed, neuron reuse, and behavior with small data.
We provide visualizations of individual neurons of a deep image recognition network during the temporal process of transfer learning. These visualizations qualitatively demonstrate various novel properties of the transfer learning process regarding the speed and characteristics of adaptation, neuron reuse, spatial scale of the represented image features, and behavior of transfer learning to small data. We publish the large-scale dataset that we have created for the purposes of this analysis.