Learning Classifiers from Synthetic Data Using a Multichannel Autoencoder
By Xi Zhang, Yanwei Fu, Andi Zang, Leonid Sigal, Gady Agam. 2015
43 citations recorded by the source corpus.
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.
Subfields: Artificial Intelligence, Generative Models
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Atilim Gunes Baydin
- Tuan Anh Le
- Frank D. Wood
- Sebastien C. Wong
- Victor Stamatescu
- Adam Gatt
- Mattias Kuosku
- Simon Kardell
In Generative Models
- Tuan Anh Le
- Frank D. Wood
- Atilim Gunes Baydin
- Tomas Pfister
- Oncel Tuzel
- J. Susskind
- Diane Joyce Cook
- Garrett Wilson
Citations in the corpus
- Learning from Simulated and Unsupervised Images through Adversarial Training (2016, background): “[46] learn a Multichannel Autoencoder to reduce the domain shift between real and synthetic data.”
- Understanding Data Augmentation for Classification: When to Warp? (2016, background): “This approach was investigated by [13], in creating synthetic images of building roofs to augment real image data.”
- Inference Compilation and Universal Probabilistic Programming (2016, background): “The fact that we had to tune our priors highlights the issues of model bias and “synthetic gap” (Zhang et al., 2015) when training models with synthetic data and testing with real data.”
- Autonomous vehicle control via deep reinforcement learning (2017, background): “Due to this, the use of synthetic data for training machine learning algorithms have been an active research field [8][9][10].”