InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
By Xi Chen, Yan Duan, Rein Houthooft, John Schulman, Ilya Sutskever, Pieter Abbeel. 2016
4,557 citations recorded by the source corpus.
This paper describes InfoGAN, an information-theoretic extension to the Generative Adversarial Network that is able to learn disentangled representations in a completely unsupervised manner. InfoGAN is a generative adversarial network that also maximizes the mutual information between a small subset of the latent variables and the observation. We derive a lower bound to the mutual information objective that can be optimized efficiently, and show that our training procedure can be interpreted as a variation of the Wake-Sleep algorithm. Specifically, InfoGAN successfully disentangles writing styles from digit shapes on the MNIST dataset, pose from lighting of 3D rendered images, and background digits from the central digit on the SVHN dataset. It also discovers visual concepts that include hair styles, presence/absence of eyeglasses, and emotions on the CelebA face dataset. Experiments show that InfoGAN learns interpretable representations that are competitive with representations learned by existing fully supervised methods.
Subfields: Computer Vision and Pattern Recognition, Generative Models, Representation & Self-Supervised Learning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Trevor Darrell
- Xiaogang Wang
- Jitendra Malik
- Andrea Vedaldi
- Andrew Zisserman
In Generative Models
- Stefano Ermon
- Jun-Yan Zhu
- Yoshua Bengio
- Richard Zhang
- Eli Shechtman
- Arash Vahdat
- Jiaming Song
- Fei Ye
In Representation & Self-Supervised Learning
- Yoshua Bengio
- Bernhard Schölkopf
- Sergey Levine
- Trevor Darrell
- Andrea Vedaldi
- Dacheng Tao
- Francesco Locatello
- J. Urgen Schmidhuber
Citations in the corpus
- CTRL: A Conditional Transformer Language Model for Controllable Generation (2019, methodology): “Much research then focused on methods for controlling the generation process and improving estimation of generative distributions (Arjovsky et al., 2017; Chen et al., 2016; Kingma & Welling, 2013).”
- A Sober Look at the Unsupervised Learning of Disentangled Representations and their Evaluation (2020, background): “They should contain all the information present in x in a compact and interpretable structure (Bengio et al., 2013; Kulkarni et al., 2015; Chen et al., 2016) while being independent from the task at hand (Goodfellow et al.”
- Are GANs Created Equal? A Large-Scale Study (2017, methodology): “To ensure a fair comparison, we made the following choices: (i) we use the generator and discriminator architecture from INFO GAN [6] as the resulting function space is rich enough and all considered GANs were not originally designed for this architecture.”
- Neural Discrete Representation Learning (2017, background): “Learning representations with continuous features have been the focus of many previous work [16, 39, 6, 9] however we concentrate on discrete representations [27, 33, 8, 28] which are potentially a more natural fit for many of the modalities we are interested in. Language is inherently discrete,…”
- A Style-Based Generator Architecture for Generative Adversarial Networks (2018, background): “While it is possible to add an extra network for this purpose [7, 11, 14], we want to avoid investing effort into a component that is not a part of the actual solution.”
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (2018, background): “This is inspired by Chen et al. (2016), and is chosen to allow some factors of variation to be discrete, while others are continuous.”
- Autoregressive Quantile Networks for Generative Modeling (2018, background): “These have led to breakthroughs in state-of-the-art generation of natural images (Karras et al., 2017) and audio (van den Oord et al., 2016a), and even been used for unsupervised learning of disentangled representations (Hig-gins et al., 2017; Chen et al., 2016).”
- Generative Adversarial Networks (2021, neutral): “Several strategies have been developed to address mode collapse, s.a. adding noise to the input of the Generator Network , using feature matching [14], or using different architectures for the Generator Network and Discriminator Network , s.a. the CycleGAN [15].”
- Dual Contradistinctive Generative Autoencoder (2020, positive): “InfoGAN [9] aims to maximize mutual information between latent variables and the output of the generator.”
- Diagnosing and Enhancing VAE Models (2019, methodology): “…henceforth denoted as 2-Stage VAE , against three baseline VAE models differing only in the decoder output layer: a Gaussian layer with fixed γ , a Gaussian layer with a learned γ , and a cross-entropy layer as has been adopted in several previous applications involving images (Chen et al., 2016b).”
- Unrolled Generative Adversarial Networks (2016, methodology): “…the GAN framework has also enabled a number of successful extensions of the technique, for instance for structured prediction (Reed et al., 2016a;b; Odena et al., 2016), training energy based models (Zhao et al., 2016), and combining the GAN loss with a mutual information loss (Chen et al., 2016).”
- Learning deep representations by mutual information estimation and maximization (2018, background): “…bi-directional adversarial models (Dumoulin et al., 2016; Donahue et al., 2016), which adversarially train an encoder / decoder to match their respective joint distributions (which typically increases the marginal entropy in Equation 1) or to minimize the reconstruction error (Chen et al., 2016).”
- Unsupervised Image-to-Image Translation Networks (2017, background): “InfoGAN (Chen et al., 2016) learned to a more interpretable latent representation.”
- Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis (2017, background): “It has also been empirically shown [2–4, 8] that adding auxiliary tasks seems to reduce such a tendency and improve convergence stability; side information is therefore often used as either additional information to the discriminator / generator [4, 5, 8], as a quantity to be reconstructed by the di”
- Learning from Simulated and Unsupervised Images through Adversarial Training (2016, background): “[2] propose InfoGAN, an information-theoretic extension of GAN, that allows learning of meaningful representations.”