Neural Network model for the CiteULike dataset using PyTorch. The model utilizes negative sampling, as the dataset only contains positive samples.
A simple model that can be used with the CiteULike dataset. It achieves roughly 90% accuracy after 15 epochs. As the original dataset only contains positive samples, the data generation adds an equal amount of random negative samples for every user, meaning that if user A has interacted with 20 articles, 20 random articles are drawn and put in the dataset as negative samples. Validation set only contains positive samples for obvious reasons.
To run this program you simply need a python 3.x (that works with PyTorch).
pip install -r requirements.txt
The dataset is included in the dataset folder, however you need to generate the training and validation set first.
Do the following in a python console:
import pandas as pd
import numpy as np
from random import randint
import os
dataset = os.path.join('dataset')
from data import to_csv_citeulike
to_csv_citeulike()
python main.py
You can play around with the hyperparameters in the main.py or the model in model.py
The package is Open Source Software released under the MIT license.