Deep Variational Canonical Correlation Analysis

Abstract

We present deep variational canonical correlation analysis (VCCA), a deep multi-view learning model that extends the latent variable model interpretation of linear CCA to nonlinear observation models parameterized by deep neural networks. We derive variational lower bounds of the data likelihood by parameterizing the posterior probability of the latent variables from the view that is available at test time. We also propose a variant of VCCA called VCCA-private that can, in addition to the "common variables" underlying both views, extract the "private variables" within each view, and disentangles the shared and private information for multi-view data without hard supervision. Experimental results on real-world datasets show that our methods are competitive across domains.

References in corpus (7)

• Stochastic Backpropagation and Approximate Inference in Deep Generative Models
• Variational Inference with Normalizing Flows
• InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
• DRAW: A Recurrent Neural Network For Image Generation
• A kernel method for canonical correlation analysis
• Deep Variational Information Bottleneck
• Adversarial Autoencoders

Cited by in corpus (33)

• Self-Supervised Speech Representation Learning: A Review
• Deep Variational Information Bottleneck
• Multimodal Generative Models for Scalable Weakly-Supervised Learning
• Learning Robust Representations via Multi-View Information Bottleneck
• Semi-supervised Deep Generative Modelling of Incomplete Multi-Modality Emotional Data
• MISA: Modality-Invariant and -Specific Representations for Multimodal Sentiment Analysis
• Trusted Multi-View Classification
• Canonical Correlation Analysis (CCA) Based Multi-View Learning: An Overview
• Gaussian Process Prior Variational Autoencoders
• Quantum-inspired canonical correlation analysis for exponentially large dimensional data
• Multimodal Variational Autoencoders for Semi-Supervised Learning: In Defense of Product-of-Experts
• Multimodal Generative Models for Compositional Representation Learning
• Disentangled Cumulants Help Successor Representations Transfer to New Tasks
• Sparse Canonical Correlation Analysis via Concave Minimization
• Variational Deep Semantic Hashing for Text Documents
• Acoustic Feature Learning via Deep Variational Canonical Correlation Analysis
• Semi-supervised Bayesian Deep Multi-modal Emotion Recognition
• Shared Generative Latent Representation Learning for Multi-view Clustering
• Deep Tensor CCA for Multi-view Learning
• Revisiting Deep Generalized Canonical Correlation Analysis
• Deep Co-Attention Network for Multi-View Subspace Learning
• Variational Inference for Deep Probabilistic Canonical Correlation Analysis
• Multi-view Temporal Alignment for Non-parallel Articulatory-to-Acoustic Speech Synthesis
• Acoustic feature learning using cross-domain articulatory measurements
• A Generative Map for Image-based Camera Localization
• ORCCA: Optimal Randomized Canonical Correlation Analysis
• Disentangled Variational Information Bottleneck for Multiview Representation Learning
• Multi-view Alignment and Generation in CCA via Consistent Latent Encoding
• Tensor Canonical Correlation Analysis with Convergence and Statistical Guarantees
• Learning relevant features for statistical inference
• Adversarial Canonical Correlation Analysis
• Variational learning across domains with triplet information
• Learning Disentangled Latent Factors from Paired Data in Cross-Modal Retrieval: An Implicit Identifiable VAE Approach