Does End-to-End Trained Deep Model Always Perform Better than Non-End-to-End Counterpart?
Ikuro Sato, Guoqing Liu, Kohta Ishikawa, Teppei Suzuki, Masayuki Tanaka
Abstract
Ikuro Sato, Guoqing Liu, Kohta Ishikawa, Teppei Suzuki, Masayuki Tanaka
Abstract
It has been rigorously demonstrated that an end-to-end (E2E) differentiable formulation of a deep neural network can turn a complex recognition problem into a unified optimization task that can be solved by some gradient descent method. Although E2E network optimization yields a powerful fitting ability, the joint optimization of layers is known to potentially bring situations where layers co-adapt one to the other in a complex way that harms generalization ability. This work aims to numerically evaluate the generalization ability of a particular non-E2E network optimization approach known as FOCA (Feature-extractor Optimization through Classifier Anonymization) that helps to avoid complex co-adaptation, with careful hyperparameter tuning. In this report, we present intriguing empirical results where the non-E2E trained models consistently outperform the corresponding E2E trained models on three image-classification datasets. We further show that E2E network fine-tuning, applied after the feature-extractor optimization by FOCA and the following classifier optimization with the fixed feature extractor, indeed gives no improvement on the test accuracy. The source code is available at https://github.com/DensoITLab/FOCA-v1.
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
It has been rigorously demonstrated that an end-to-end (E2E) differentiable formulation of a deep neural network can turn a complex recognition problem into a unified optimization task that can be solved by some gradient descent method. Although E2E network optimization yields a powerful fitting ability, the joint optimization of layers is known to potentially bring situations where layers co-adapt one to the other in a complex way that harms generalization ability. This work aims to numerically evaluate the generalization ability of a particular non-E2E network optimization approach known as FOCA (Feature-extractor Optimization through Classifier Anonymization) that helps to avoid complex co-adaptation, with careful hyperparameter tuning. In this report, we present intriguing empirical results where the non-E2E trained models consistently outperform the corresponding E2E trained models on three image-classification datasets. We further show that E2E network fine-tuning, applied after the feature-extractor optimization by FOCA and the following classifier optimization with the fixed feature extractor, indeed gives no improvement on the test accuracy. The source code is available at https://github.com/DensoITLab/FOCA-v1.
Key concepts: End-to-end principle, End of history, Dead end, End milling, End user, Computer science, End point, Front and back ends