A Heat Diffusion Perspective on Geodesic Preserving Dimensionality Reduction

Authors: Guillaume Huguet, Alexander Tong, Edward De Brouwer, Yanlei Zhang, Guy Wolf, Ian Adelstein, Smita Krishnaswamy

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Results show that our method outperforms existing state-of-the-art in preserving ground truth manifold distances, and preserving cluster structure in toy datasets. We also showcase our method on single cell RNA-sequencing datasets with both continuum and cluster structure, where our method enables interpolation of withheld timepoints of data.
Researcher Affiliation Academia Guillaume Huguet1 Alexander Tong1 Edward De Brouwer2 Yanlei Zhang1 Guy Wolf1 Ian Adelstein2 Smita Krishnaswamy2 1Université de Montréal; Mila Quebec AI Institute 2 Yale University
Pseudocode Yes In Alg. 1, we present the main steps of our algorithm using the heat-geodesic dissimilarity. A detailed version is presented in Appendix A.
Open Source Code Yes 3https://github.com/KrishnaswamyLab/HeatGeo
Open Datasets Yes We use the PBMC dataset, the Swiss roll, the Tree dataset, MNIST [8], and COIL-20 [23] dataset. ... and two from the 2022 NeurIPS multimodal single-cell integration challenge4. 4https://www.kaggle.com/competitions/open-problems-multimodal/
Dataset Splits Yes For all models, we perform sample splitting with a 50/50 validation-test split. The validation and test sets each consists of 5 repetitions with different random initializations.
Hardware Specification Yes The experiments were performed on a compute node with 16 Intel Xeon Platinum 8358 Processors and 64GB RAM.
Software Dependencies No The paper mentions software like 'Pygsp' and 'Scanpy' but does not provide specific version numbers for these or any other key software dependencies.
Experiment Setup Yes In Table 5, we report the values of hyperparameters used to compute the different embeddings.