Read the original document by opening this link in a new tab.
Table of Contents
1. INTRODUCTION
2. PROBLEM FORMULATION AND TENSOR FACTORIZATION (TF)
3. NEURAL NETWORK BASED TENSOR FACTORIZATION FRAMEWORK
3.1 General Framework
3.2 Modeling Temporal Dynamics via LSTM
3.3 Fusion of LSTM and TF
3.4 Learning Process
3.4.1 The Objective Function
3.4.2 Batch Normalization
4. EVALUATION
Summary
Neural Tensor Factorization is a paper that presents a model for predictive tasks on dynamic relational data. It addresses the limitations of existing models by proposing a Neural network based Tensor Factorization (NTF) model. The model leverages the long short-term memory architecture and incorporates the multi-layer perceptron structure to improve rating prediction and link prediction on dynamic relational data. The paper introduces the problem formulation, the conventional tensor factorization model, and discusses the challenges it faces. It then details the NTF framework, including the fusion of LSTM and TF, the learning process, and the application of batch normalization. The effectiveness of NTF is demonstrated through experiments on rating prediction and link prediction tasks.