https://github.com/fandreuz/parallel-mapped-distance-matrix
Parallel mapped distance matrix with NumPy and Numba
Science Score: 13.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
○.zenodo.json file
-
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (6.1%) to scientific vocabulary
Keywords
Repository
Parallel mapped distance matrix with NumPy and Numba
Basic Info
Statistics
- Stars: 3
- Watchers: 1
- Forks: 0
- Open Issues: 0
- Releases: 0
Topics
Metadata Files
README.md
Parallel MDM
Mapped distance matrix
The Mapped Distance Matrix (MDM) of two sets $\mathcal{X}, \mathcal{Y}$ of n-dimensional points is an algebraic structure which is defined in general as follows, given a mapping $f$:
$$\mathbf{M}(\mathcal{X}, \mathcal{Y}, f)_{i,j} := f(\Vert \mathcal{X}_i - \mathcal{Y}_j\Vert)$$
where $\Vert \cdot \Vert$ is an appropriate distance notion on the space of definition of $\mathcal{X}$ and $\mathcal{Y}$.
The problem might be augmented by weighting the contributions with a matrix of weights $\mathbf{W}$; the updated definition is then:
$$\mathbf{M}(\mathcal{X}, \mathcal{Y}, f)_{i,j} := \mathbf{W}_{i,j} f(\Vert \mathcal{X}_{i} - \mathcal{Y}_{j}\Vert)$$
A particularly popular form of the problem (which is also what we treat in this repository) occurs when weights are defined individually for the members of $\mathcal{Y}$ (i.e. the columns of $\mathbf{W}$ are taken constants):
$$\mathbf{M}(\mathcal{X}, \mathcal{Y}, f)_{i,j} := \mathbf{W}_{j} f(\Vert \mathcal{X}_{i} - \mathcal{Y}_{j}\Vert)$$
A notable case: uniform grid
In general $\mathcal{X}, \mathcal{Y}$ identify two general sets of points. A few applications allow more assumptions on the two sets. For instance, $\mathcal{X}$ might be taken to be an uniform grid. In this case a few interesting optimizization can be taken into account for the computation of the matrix.
More assumptions
Practical applications usually require huge sets of points, which causes memory errors on commonly used devices. This is why it's preferrable to compute the vector $\tilde{\mathbf{M}}$ defined below instead of $\mathbf{M}$:
$$\tilde{\mathbf{M}}_{i} := \sum_{j} \mathbf{M}_{i,j}$$
For most use cases this is enough.
Roadmap
- Algorithms
- [x] Uniform grid algorithm
- [x] Scattered points algorithm
- [ ] Fourier-transfor based algorithm
- [ ] Backends
- [ ] NumPy/Numba
- [ ] PyTorch
- [ ] JAX(?)
- [ ] Parallelization
- [x] Multithreading/Multiprocessing
- [ ] GPU w/ PyTorch
- [ ] GPU w/ JAX
- [ ] CUDA kernels(?)
- [ ] Tests
- [ ] Documentation
- [ ] Benchmark (+comparison with competitors)
- [ ] CPU
- [ ] GPU
- [ ] Several different bin sizes
- [ ]
pts_per_future
- [ ] Future
- [ ] Periodicity
- [ ] More general about distance definitions
Owner
- Name: Francesco Andreuzzi
- Login: fandreuz
- Kind: user
- Location: Geneva, Switzerland
- Company: CERN
- Repositories: 1
- Profile: https://github.com/fandreuz
CSE MSc student | SWE @cern
GitHub Events
Total
- Push event: 1
Last Year
- Push event: 1
Issues and Pull Requests
Last synced: 9 months ago
All Time
- Total issues: 0
- Total pull requests: 4
- Average time to close issues: N/A
- Average time to close pull requests: 14 minutes
- Total issue authors: 0
- Total pull request authors: 1
- Average comments per issue: 0
- Average comments per pull request: 0.0
- Merged pull requests: 4
- Bot issues: 0
- Bot pull requests: 0
Past Year
- Issues: 0
- Pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Issue authors: 0
- Pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0
Top Authors
Issue Authors
Pull Request Authors
- fandreuz (4)