Papers

Related modeling papers

1. Event segmentation

   • Hierarchical Multiscale Recurrent Neural Networks [code]
   • Hierarchical Boundary-Aware Neural Encoder for Video Captioning [code]
   • Modeling Unsupervised Event Segmentation: Learning Event Boundaries from Prediction Errors
   • TRACX: A Recognition-Based Connectionist Framework for Sequence Segmentation and Chunk Extraction

2. Question answering

   • Ask Me Anything: Dynamic Memory Networks for Natural Language Processing [code]
   • Memory networks
   • Dynamic Coattention Networks For Question Answering [code]

3. Storytelling

   • Towards Story-like Visual Explanations by Watching Movies and Reading Books [Project page, blog, code]

4. Attention (temporal)

   • Describing Videos by Exploiting Temporal Structure [code]
   • Attention Is All You Need

5. External memory

   • Hybrid computing using a neural network with dynamic external memory [code]
   • Neural Turing Machines [code]
   • A Growing Long-term Episodic & Semantic Memory

6. Statistical learning

   • Distributed representations of action sequences in anterior cingulate cortex: A recurrent neural network approach

7. Symbolic computation

   • NIPS workshop: Integrating Neural and Symbolic Approaches
   • Sentence gestalt model John and Mcclelland
   • Holographic reduced representations [HRR playground, a tensorflow example]
     • Holographic Recurrent Networks
     • Neural representations of compositional structures: representing and manipulating vector spaces with spiking neurons
     • Integrating structure and meaning: a distributed model of analogical mapping
     • Embedding Probabilities in Predication Space with Hermitian Holographic Reduced Representations
     • Associative Long Short-Term Memory [code]
   • Schema Networks: Zero-shot Transfer with a Generative Causal Model of Intuitive Physics
   • learning relations
     • Embedding Entities and Relations for Learning and Inference in Knowledge Bases
     • Learning Multi-Relational Semantics Using Neural-Embedding Models
     • Dynamic Entity Representations in Neural Language Models
     • Neural Symbolic Machines: Learning Semantic Parsers on Freebase with Weak Supervision
   • learning relation between graphs
     • graph2vec: Learning Distributed Representations of Graphs
   • learning grammar
     • Recurrent Neural Network Grammars [code, motivation, sup]
     • Jointly Learning Sentence Embeddings and Syntax with Unsupervised Tree-LSTMs
     • Improved Semantic Representations From Tree-Structured LSTMs
     • Cross-Sentence N-ary Relation Extraction with Graph LSTMs
   • high level task
     • Learning to Compose Words into Sentences with Reinforcement Learning

8. Schema representation:

   • Neural Episodic Control

9. Empirical papers:

   • Rethinking the Role of the Angular Gyrus in Remembering the Past and Imagining the Future: The Contextual Integration Model

10. Other stuff

    • Imagination-Augmented Agents for Deep Reinforcement Learning - Combine model-free and model-based learning.
    • Overcoming catastrophic forgetting in neural networks [code, blog]- Learn multiple tasks sequentially
    • One Model To Learn Them All [code] - Learn multiple tasks; Block modeling
    • Deep Residual Learning for Image Recognition [Tutorial, code] - Train deeper models.
      • Densely Connected Convolutional Networks [code] - Train deeper models.
    • Papers - Architecture Search