Source Code generated during Masters Studies Thesis: Exploiting Inter-session Dynamics for Long Intra-Session Sequences of Interactions with Deep Reinforcement Learning for Session-Aware Recommendation link
This repository includes a Pytorch implementation of the hgru4rec architecture as close as possible to the original implementation and an adaptation of the Deep Deterministic Policy Gradient Algorithm to adjust the hidden state of the intra-session level network by introducing inter-session based perturbations.
- Cuda 10.1 (Driver 430.64 Ubuntu 16.04)
- Pytorch 1.3.1
- Pandas 0.24
- PyTables 3.4.4
- Numpy 1.15.4
Additionally you will need to create several folders inside the root folder to store models and preprocessed data
- Datasets
- Models
- Results
This repository includes preprocessing files for the following datasets:
- 30 Music Dataset
- Repeat buyers challenge
To build datasets execute
/data/BuildDataset.ipynb
In your terminal execute the file adding the split.sh