Semi-Supervised Learning with Normalizing Flows

By Pavel Izmailov, Polina Kirichenko, Marc Finzi, Andrew Gordon Wilson. 2019

137 citations recorded by the source corpus.

Normalizing flows transform a latent distribution through an invertible neural network for a flexible and pleasingly simple approach to generative modelling, while preserving an exact likelihood. We propose FlowGMM, an end-to-end approach to generative semi supervised learning with normalizing flows, using a latent Gaussian mixture model. FlowGMM is distinct in its simplicity, unified treatment of labelled and unlabelled data with an exact likelihood, interpretability, and broad applicability beyond image data. We show promising results on a wide range of applications, including AG-News and Yahoo Answers text data, tabular data, and semi-supervised image classification. We also show that FlowGMM can discover interpretable structure, provide real-time optimization-free feature visualizations, and specify well calibrated predictive distributions.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Probabilistic & Bayesian Methods, Signal Processing

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

1. Arash Vahdat
2. Pavel Izmailov
3. Andrew Gordon Wilson
4. Polina Kirichenko
5. Karsten Kreis
6. Jan Kautz
7. Alexander Gerhard Schwing
8. Jyoti Aneja

In Computer Vision and Pattern Recognition

1. Arash Vahdat
2. Jan Kautz
3. Karsten Kreis
4. Luc Van Gool
5. Ming-Hsuan Yang
6. Alexander Gerhard Schwing
7. Dacheng Tao
8. Trevor Darrell

In Probabilistic & Bayesian Methods

1. Pavel Izmailov
2. Andrew Gordon Wilson
3. Polina Kirichenko
4. Jan Kautz
5. Karsten Kreis
6. Arash Vahdat
7. Zhisheng Xiao
8. Alexander Gerhard Schwing

In Signal Processing

1. Jan Kautz
2. Karsten Kreis
3. Zhisheng Xiao
4. Andrew Gordon Wilson
5. Alexander Gerhard Schwing
6. Jyoti Aneja
7. Ashish Khetan
8. Arash Vahdat

Citations in the corpus

• VAEBM: A Symbiosis between Variational Autoencoders and Energy-based Models (2020, background): “…distribution alignment across domains (Zhu et al., 2017; Liu et al., 2017; Tzeng et al., 2017) and semi-supervised learning (Kingma et al., 2014; Izmailov et al., 2020) to 3D point cloud generation (Yang et al., 2019), light-transport simulation (Müller et al., 2019), molecular modeling…”
• Task-agnostic Continual Learning with Hybrid Probabilistic Models (2021, unknown): “Normalizing flows have been shown to perform well on classification and out-of-distribution detection tasks using the image embeddings [Izmailov et al., 2020, Kirichenko et al., 2020, Zhang et al., 2020, Zisselman and Tamar, 2020].”
• NCP-VAE: Variational Autoencoders with Noise Contrastive Priors (2020, background): “…2020), speech generation (Oord et al., 2016; Ping et al., 2020), image captioning (Aneja et al., 2019; Deshpande et al., 2019; Aneja et al., 2018), semi-supervised learning (Kingma et al., 2014; Izmailov et al., 2020), and representation learning (Van Den Oord et al., 2017; Fortuin et al., 2018).”
• TabTransformer: Tabular Data Modeling Using Contextual Embeddings (2020, methodology): “(Izmailov et al. 2019) give a semi-supervised method based on density estimation and evaluate their approach on tabular data.”
• Why Normalizing Flows Fail to Detect Out-of-Distribution Data (2020, methodology): “Connection with prior work Flows can be used as classifiers separating different classes of the same dataset [28, 19, 1], which further highlights the fact that flows can distinguish images based on their contents when trained to do so.”