COTIC

Description

In this project, the problem of return time prediction and event time prediction is considered. To solve this problem, four models were considered: Continuous Time CNN, Transformer Hawkes, WaveNet and WaveNet with intensity interpolation. Here is a convenient pipeline for working with them and launching them. To get acquainted with the structure of the code and the features, you can look at the information provided below.

Architecture of model

• Continuous time CNN

• Transformer Hawkes

Datasets

• LinkedIn
• Amazon
• IPTV
• Synthetic Hawkes processes

The datasets are taken from cloud drive.

How to run

Install dependencies

# clone project
git clone git@github.com:VladislavZh/CCNN.git
cd CCNN

# [OPTIONAL] create conda environment
conda create -n myenv python=3.9
conda activate myenv

# install pytorch according to instructions
# https://pytorch.org/get-started/

# install requirements
pip install -r requirements.txt

Train model with default configuration

# train on CPU
python train.py trainer.gpus=0

# train on GPU
python train.py trainer.gpus=1

Train different model architectures with comet logger:

Change api_key, project_name, workspace in configs/logger/comet.yaml

python train.py --config-name=train.yaml data_dir=/content/data_utils/ model.net.num_types=count_of_event_type

data_dir should be a folder with csv files where each file represent a sequence with event and timestamp.
--config-name one of model from configs/

You can override any parameter from command line like this

python train.py trainer.max_epochs=20 datamodule.batch_size=64

Project Structure

The directory structure of new project looks like this:

├── configs                   <- Hydra configuration files
│   ├── callbacks                <- Callbacks configs
│   ├── datamodule               <- Datamodule configs
│   ├── debug                    <- Debugging configs
│   ├── experiment               <- Experiment configs
│   ├── hparams_search           <- Hyperparameter search configs
│   ├── local                    <- Local configs
│   ├── log_dir                  <- Logging directory configs
│   ├── logger                   <- Logger configs
│   ├── model                    <- Model configs
│   ├── trainer                  <- Trainer configs
│         │     
│         ├── test.yaml             <- Main config for testing
│         ├── train.yaml            <- Main config for training
│         ├── ...
├── data                   <- Project data
│
├── logs                   <- Logs generated by Hydra and PyTorch Lightning loggers
│
├── notebooks              <- Jupyter notebooks. Naming convention is a number (for ordering),
│                             the creator's initials, and a short `-` delimited description,
│                             e.g. `1.0-jqp-initial-data-exploration.ipynb`.
│
├── scripts                <- Shell scripts
│
├── src                    <- Source code
│   ├── datamodules              <- Lightning datamodules
│   ├── metrics                  <- Metrics for different model
│   ├── models                   <- Lightning models
│   ├── utils                    <- Utility scripts
│   ├── vendor                   <- Third party code that cannot be installed using PIP/Conda
│   │
│   ├── testing_pipeline.py
│   └── training_pipeline.py
│
├── tests                  <- Tests of any kind
│   ├── helpers                  <- A couple of testing utilities
│   ├── shell                    <- Shell/command based tests
│   └── unit                     <- Unit tests
│
├── test.py               <- Run testing
├── train.py              <- Run training
│
├── .env.example              <- Template of the file for storing private environment variables
├── .gitignore                <- List of files/folders ignored by git
├── .pre-commit-config.yaml   <- Configuration of pre-commit hooks for code formatting
├── requirements.txt          <- File for installing python dependencies
├── setup.cfg                 <- Configuration of linters and pytest
└── README.md

References

• The Neural Hawkes Process: A Neurally Self-Modulating Multivariate Point Process
• Continuous Cnn For Nonuniform Time Series