Variational Autoencoder for Deep Learning of Images, Labels and Captions
arXiv:1609.08976
Abstract
A novel variational autoencoder is developed to model images, as well as associated labels or captions. The Deep Generative Deconvolutional Network (DGDN) is used as a decoder of the latent image features, and a deep Convolutional Neural Network (CNN) is used as an image encoder; the CNN is used to approximate a distribution for the latent DGDN features/code. The latent code is also linked to generative models for labels (Bayesian support vector machine) or captions (recurrent neural network). When predicting a label/caption for a new image at test, averaging is performed across the distribution of latent codes; this is computationally efficient as a consequence of the learned CNN-based encoder. Since the framework is capable of modeling the image in the presence/absence of associated labels/captions, a new semi-supervised setting is manifested for CNN learning with images; the framework even allows unsupervised CNN learning, based on images alone.
NIPS 2016 (To appear)
References in corpus (5)
Cited by in corpus (36)
- In Defense of Pseudo-Labeling: An Uncertainty-Aware Pseudo-label Selection Framework for Semi-Supervised Learning
- Topic-Guided Variational Autoencoders for Text Generation
- 6GCVAE: Gated Convolutional Variational Autoencoder for IPv6 Target Generation
- EM-like Learning Chaotic Dynamics from Noisy and Partial Observations
- Neural Language Generation: Formulation, Methods, and Evaluation
- Spatial-Phase Shallow Learning: Rethinking Face Forgery Detection in Frequency Domain
- RETRIEVE: Coreset Selection for Efficient and Robust Semi-Supervised Learning
- Semi-supervised learning based on generative adversarial network: a comparison between good GAN and bad GAN approach
- Regularizing linear inverse problems with convolutional neural networks
- Statistical Inference in Mean-Field Variational Bayes
- Blind Image Denoising and Inpainting Using Robust Hadamard Autoencoders
- Multi-Task Curriculum Framework for Open-Set Semi-Supervised Learning
- Variational Autoencoders for Jet Simulation
- Label Propagation Adaptive Resonance Theory for Semi-supervised Continuous Learning
- Modality Conversion of Handwritten Patterns by Cross Variational Autoencoders
- Gaussian Smoothen Semantic Features (GSSF) -- Exploring the Linguistic Aspects of Visual Captioning in Indian Languages (Bengali) Using MSCOCO Framework
- DeltaGAN: Towards Diverse Few-shot Image Generation with Sample-Specific Delta
- F2GAN: Fusing-and-Filling GAN for Few-shot Image Generation
- Semi-supervised Learning using Adversarial Training with Good and Bad Samples
- Generating various airfoil shapes with required lift coefficient using conditional variational autoencoders
- General-Purpose User Embeddings based on Mobile App Usage
- Evaluating for Diversity in Question Generation over Text
- Anatomy-Aware Cardiac Motion Estimation
- Autoencoding Improves Pre-trained Word Embeddings
- Regularized Sequential Latent Variable Models with Adversarial Neural Networks
- Sum-Product-Attention Networks: Leveraging Self-Attention in Probabilistic Circuits
- That looks interesting! Personalizing Communication and Segmentation with Random Forest Node Embeddings
- Loosely Coupled Federated Learning Over Generative Models
- Learning Sparsity of Representations with Discrete Latent Variables
- Tackling Long-Tailed Relations and Uncommon Entities in Knowledge Graph Completion
- Image to Language Understanding: Captioning approach
- Enhanced Variational Inference with Dyadic Transformation
- Learning a Generative Model of Cancer Metastasis
- Decoupled Variational Embedding for Signed Directed Networks
- Towards Stable Imbalanced Data Classification via Virtual Big Data Projection
- Learning more expressive joint distributions in multimodal variational methods