RecNet

Session based recommendation system using state of the art Transformers and Matrix Factorization on MovieLens-1M.
Please refer to the Report to view a detailed explanation of our project.

Model Architecture

Results

Learned Item Embeddings from Scratch

Install Dependencies

pip install -r requirements.txt

Instructions to run code

• Create save/ directory inside the ./Code folder in order to save checkpoints
• Run python Code/train.py --max_seq_len 200 --num_layers 2 for the Transformer Model
• Run python Code/trainRNN.py --max_seq_len 200 --num_layers 2 for the RNN Model

Tensorboard

tensorboard --logdir=runs