Poster
Deep Model Transferability from Attribution Maps
Jie Song · Yixin Chen · Xinchao Wang · Chengchao Shen · Mingli Song

Wed Dec 11th 10:45 AM -- 12:45 PM @ East Exhibition Hall B + C #100

Exploring the transferability between heterogeneous tasks sheds light on their intrinsic interconnections, and consequently enables knowledge transfer from one task to another so as to reduce the training effort of the latter. In this paper, we propose an embarrassingly simple yet very efficacious approach to estimating the transferability of deep networks, especially those handling vision tasks. Unlike the seminal work of \emph{taskonomy} that relies on a large number of annotations as supervision and is thus computationally cumbersome, the proposed approach requires no human annotations and imposes no constraints on the architectures of the networks. This is achieved, specifically, via projecting deep networks into a \emph{model space}, wherein each network is treated as a point and the distances between two points are measured by deviations of their produced attribution maps. The proposed approach is several-magnitude times faster than taskonomy, and meanwhile preserves a task-wise topological structure highly similar to the one obtained by taskonomy. Code is available at \url{https://github.com/zju-vipa/TransferbilityFromAttributionMaps}.

Author Information

Jie Song (Zhejiang University)
Yixin Chen (Zhejiang University)
Xinchao Wang (Stevens Institute of Technology)
Chengchao Shen (Zhejiang University)
Mingli Song (Zhejiang University)

More from the Same Authors

  • 2018 Poster: Dual Swap Disentangling »
    Zunlei Feng · Xinchao Wang · Chenglong Ke · An-Xiang Zeng · Dacheng Tao · Mingli Song