deepdow (read as "wow") is a Python package connecting portfolio optimization and deep learning. Its goal is to
facilitate research of networks that perform weight allocation in one forward pass.
pip install deepdow- Getting started: https://deepdow.readthedocs.io/en/latest/auto_examples/end_to_end/getting_started.html
- Detailed documentation: https://deepdow.readthedocs.io/en/latest
- More examples: https://deepdow.readthedocs.io/en/latest/auto_examples/index.html
deepdow attempts to merge two very common steps in portfolio optimization
- Forecasting of future evolution of the market (LSTM, GARCH,...)
- Optimization problem design and solution (convex optimization, ...)
It does so by constructing a pipeline of layers. The last layer performs the allocation and all the previous ones serve as feature extractors. The overall network is fully differentiable and one can optimize its parameters by gradient descent algorithms.
- focused on active trading strategies, it only finds allocations to be held over some horizon (buy and hold)
- one implication of this is that there is no need to handle transaction costs
- a reinforcement learning framework, however one might easily reuse
deepdowlayers in other deep learning applications - a single algorithm, instead, it is a framework that allows for easy experimentation with powerful building blocks
- all layers built on
torchand fully differentiable - integrates differentiable convex optimization (
cvxpylayers) - implements clustering based portfolio allocation algorithms
- multiple dataloading strategies (
RigidDataLoader,FlexibleDataLoader) - integration with
mlflowandtensorboardvia callbacks - provides variety of losses like sharpe ratio, maximum drawdown, ...
- simple to extend and customize
- CPU and GPU support
If you use deepdow (including ideas proposed in the documentation, examples and tests) in your research please make sure to cite it.
To obtain all the necessary citing information click on the DOI badge at top of this README and you will be automatically redirected to an external website.
Note that currently we are using Zenodo.