Geomstats: A Python Package for Riemannian Geometry in Machine Learning
Authors: Nina Miolane, Nicolas Guigui, Alice Le Brigant, Johan Mathe, Benjamin Hou, Yann Thanwerdas, Stefan Heyder, Olivier Peltre, Niklas Koep, Hadi Zaatiti, Hatem Hajri, Yann Cabanes, Thomas Gerald, Paul Chauchat, Christian Shewmake, Daniel Brooks, Bernhard Kainz, Claire Donnat, Susan Holmes, Xavier Pennec
JMLR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | This paper presents the package, compares it with related libraries, and provides relevant code examples. We show that Geomstats provides reliable building blocks to both foster research in differential geometry and statistics and democratize the use of Riemannian geometry in machine learning applications. The paper's main contribution is the introduction and description of the Geomstats software package, with examples of its usage and comparisons of its features with other libraries (Table 1 and Table 2). It does not present empirical results or data analysis from experiments. |
| Researcher Affiliation | Collaboration | Nina Miolane EMAIL Nicolas Guigui EMAIL Alice Le Brigant EMAIL Johan Mathe EMAIL Benjamin Hou EMAIL Yann Thanwerdas EMAIL Stefan Heyder EMAIL Olivier Peltre EMAIL Niklas Koep EMAIL Hadi Zaatiti EMAIL Hatem Hajri EMAIL Yann Cabanes EMAIL Thomas Gerald EMAIL Paul Chauchat EMAIL Christian Shewmake EMAIL Daniel Brooks EMAIL Bernhard Kainz EMAIL Claire Donnat EMAIL Susan Holmes EMAIL Xavier Pennec EMAIL. The author list includes affiliations from academic institutions such as Stanford University, Inria, Université Paris 1, Imperial College London, and Technische Universität Ilmenau, as well as industry-related entities like Froglabs.ai and Irt-Systemx, indicating a collaboration. |
| Pseudocode | No | The paper contains Python code snippets illustrating the usage of the Geomstats library for K-means and Tangent PCA in Section 4, but these are examples of library usage rather than structured pseudocode or algorithm blocks describing a novel method. |
| Open Source Code | Yes | The source code is freely available under the MIT license at geomstats.ai. The Git Hub repository at github.com/geomstats/geomstats offers a convenient way to ask for help or request features by raising issues. |
| Open Datasets | Yes | The package comes with a visualization module to provide intuition on differential geometry (see Figure 1), and with a datasets module that provides toy data sets on manifolds. |
| Dataset Splits | No | The paper introduces a software package and provides usage examples that generate random uniform data for demonstration purposes (e.g., `sphere.random_uniform(n_samples=10)`), but it does not describe specific training, validation, or test dataset splits for experiments. |
| Hardware Specification | No | The paper describes Geomstats, an open-source Python package for computations, and focuses on its features, implementation, and comparison with other software libraries. It mentions support for different execution backends (NumPy, PyTorch, and TensorFlow) but does not provide any specific details about the hardware used for development or testing. |
| Software Dependencies | No | The paper mentions several software dependencies and backends such as Num Py, Py Torch, Tensor Flow, and Scikit-Learn API, but it does not provide specific version numbers for any of these components, which is required for reproducibility. |
| Experiment Setup | No | The paper introduces a software package and provides code snippets illustrating its usage with specific parameters for K-means (`n_clusters=4`) and Tangent PCA (`n_components=2`). However, these are examples of library usage rather than a detailed experimental setup with hyperparameters or training configurations for reproducing scientific findings. |