Learning Classifiers from Synthetic Data Using a Multichannel Autoencoder
This addresses the challenge of effectively using synthetic data to enhance classifier training, though it appears incremental as it builds on existing autoencoder methods for data integration.
The paper tackles the problem of distribution shift between synthetic and real data for classifier learning by proposing a Multichannel Autoencoder (MCAE) to bridge this gap and jointly learn from both data types, resulting in improved feature representation for classification as validated on two datasets.
We propose a method for using synthetic data to help learning classifiers. Synthetic data, even is generated based on real data, normally results in a shift from the distribution of real data in feature space. To bridge the gap between the real and synthetic data, and jointly learn from synthetic and real data, this paper proposes a Multichannel Autoencoder(MCAE). We show that by suing MCAE, it is possible to learn a better feature representation for classification. To evaluate the proposed approach, we conduct experiments on two types of datasets. Experimental results on two datasets validate the efficiency of our MCAE model and our methodology of generating synthetic data.