Traja is a research library. Functionality must therefore be both cutting-edge and reliable. Traja is part of a wider project to increase collaboration in research, through the adoption of open-source contribution models. It is our hope that traja remains accessible to researchers and help them do higher-quality research.
Traja is currently undergoing active development with approximately 60 % of features present. Significant interface changes are still possible, however we avoid these unless absolutely necessary.
The work is currently focussed on reaching version 1.0 with feature completeness and 95 % test coverage.
The following features are required for feature completeness:
- Latent space visualisers
- Eigenspace-based
- Colour-coded to visualise evolution over time
- Delay coordinate embeddings
- State-space visualisers
- Additional encoder and decoder options in AE and VAE models
- MLP
- 1d convolution
- Pituitary gland example dataset
- Regression output visualisers
- VAE GAN models
- Additional VAE latent-space shapes
- Uniform
- A shape that works for periodic trajectories (Torus?)
- Delay coordinate embeddings
- Persistent homology diagrams of the embeddings
- Automatic code formatter
- Tutorials
- Find time of day based on activity
- Recover parameters from Pituitary ODE
- Predict stock prices with LSTMs
Traja welcomes contributions! To get started, pick up any issue labeled with good first issue! Alternatively you can read some background material or try a tutorial.
Since Traja is a library, we strive for sensible tests achieving a high level of code coverage. Future commits are required to maintain or improve code quality, test coverage. To aid in this, Travis runs automated tests on each pull request. Additionally, we run codecov on PRs.
This is a collection of papers and resources that explain the main problems we are working on with Traja.
Analysis of mice that have suffered a stroke:
- @article{10.3389/fnins.2020.00518,
- author={Justin Shenk and
- Klara J. Lohkamp and Maximilian Wiesmann and Amanda J. Kiliaan},
title={Automated Analysis of Stroke Mouse Trajectory Data With Traja}, journal={Frontiers in Neuroscience}, volume={14}, pages={518}, year={2020}, url={https://www.frontiersin.org/article/10.3389/fnins.2020.00518}, doi={10.3389/fnins.2020.00518}, issn={1662-453X},
}
Understanding the parameter space of the pituitary gland ODE (https://www.math.fsu.edu/~bertram/papers/bursting/JCNS_16.pdf):
- @article{10.1007/s10827-016-0600-1,
- author = {Fletcher, Patrick and Bertram, Richard and Tabak, Joel}, title = {From Global to Local: Exploring the Relationship between Parameters and Behaviors in Models of Electrical Excitability}, year = {2016}, publisher = {Springer-Verlag}, address = {Berlin, Heidelberg}, volume = {40}, number = {3}, issn = {0929-5313}, url = {https://doi.org/10.1007/s10827-016-0600-1}, doi = {10.1007/s10827-016-0600-1}, journal = {J. Comput. Neurosci.}, month = June, pages = {331–345},
}
TODO