Adversarial Text Generation via Feature-Mover's Distance
Explore this paper's citation graph
Summary
This work proposes to improve text-generation GAN via a novel approach inspired by optimal transport, which leads to a highly discriminative critic and easy-to-optimize objective, overcoming the mode-collapsing and brittle-training problems in existing methods.
- Type
- article
- Published
- 2018-09-17
- Cited by
- 130
- References
- 67
- Access
- Open access
- OpenAlex
- https://openalex.org/W2891641674
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:52292169
Keywords
Computer science, Discriminative model, Feature (linguistics), Metric (unit), Artificial intelligence
References
- Caltech-UCSD Birds 200
- Recurrent neural network based language model
- Cryptography, information theory, and error-correction - a handbook for the 21st century
- Convolutional Neural Networks for Sentence Classification
- Show and tell: A neural image caption generator
- From captions to visual concepts and back
- Theory of Maxima and the Method of Lagrange
- Long Short-Term Memory
- Bleu: a Method for Automatic Evaluation of Machine Translation
- A metric for distributions with applications to image databases
- Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
- Sinkhorn Distances: Lightspeed Computation of Optimal Transport
- How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?
- Minimum Risk Training for Neural Machine Translation
- A Kernel Two-Sample Test
- Discriminative Regularization for Generative Models
- Sequence-to-Sequence Learning as Beam-Search Optimization
- Improved Techniques for Training GANs
- SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient
- GANS for Sequences of Discrete Elements with the Gumbel-softmax Distribution
Cited by
- A Survey of the Usages of Deep Learning for Natural Language Processing
- Generating Text through Adversarial Training Using Skip-Thought Vectors
- Learning Criteria and Evaluation Metrics for Textual Transfer between Non-Parallel Corpora
- Language GANs Falling Short
- Adversarial Discrete Sequence Generation without Explicit NeuralNetworks as Discriminators
- GAN Augmented Text Anomaly Detection with Sequences of Deep Statistics
- Topic-Guided Variational Auto-Encoder for Text Generation
- Adversarial Sub-sequence for Text Generation
- Improving Textual Network Embedding with Global Attention via Optimal Transport
- Towards Generating Long and Coherent Text with Multi-Level Latent Variable Models
- Improving Cross-lingual Entity Alignment via Optimal Transport
- How Sequence-to-Sequence Models Perceive Language Styles?
- Towards Unsupervised Image Captioning With Shared Multimodal Embeddings
- Defense Against Adversarial Attacks Using Feature Scattering-based Adversarial Training
- Unsupervised Evaluation Metrics and Learning Criteria for Non-Parallel Textual Transfer
- Rethinking Text Attribute Transfer: A Lexical Analysis
- Adversarial Learning of Deepfakes in Accounting
- The Detection of Distributional Discrepancy for Text Generation
- CatGAN: Category-aware Generative Adversarial Networks with Hierarchical Evolutionary Learning for Category Text Generation
- Mechanisms for Automatic Training Data Labeling for Machine Learning