Topologically Regularized Data Embeddings

Authors: Robin Vandaele, Bo Kang, Jefrey Lijffijt, Tijl De Bie, Yvan Saeys

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We include thorough experiments on synthetic and real data that highlight the usefulness and versatility of this approach, with applications ranging from modeling high-dimensional single-cell data, to graph embedding.
Researcher Affiliation Academia Robin Vandaele, Bo Kang, Jefrey Lijffijt, Tijl De Bie, Yvan Saeys Ghent University, Gent, Belgium
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available on github.com/robinvndaele/topembedding.
Open Datasets Yes We considered a single-cell trajectory data set of 264 cells in a 6812-dimensional gene expression space (Cannoodt et al., 2018; Saelens et al., 2019).
Dataset Splits Yes All points in the 2D embeddings were then split into 90% points for training and 10% for testing. Consecutively, we used 5-fold CV on the training data to tune the regularization hyperparameter C {1e 2, 1e 1, 1, 1e1, 1e2}.
Hardware Specification Yes Topological optimization was performed in Pytorch on a machine equipped with an Intel Core TM i7 processor at 2.6GHz and 8GB of RAM, using code adapted from Br uel-Gabrielsson et al. (2020).
Software Dependencies No The paper mentions software like Pytorch, SCIKIT-LEARN, and Dionysus but does not provide specific version numbers for these dependencies.
Experiment Setup Yes Tables 1 & 2 summarize data sizes, hyperparameters, losses, and optimization times.