• Corpus ID: 239016008

Topologically Regularized Data Embeddings

@article{Vandaele2021TopologicallyRD,
  title={Topologically Regularized Data Embeddings},
  author={Robin Vandaele and Bo Kang and Jefrey Lijffijt and Tijl De Bie and Yvan Saeys},
  journal={ArXiv},
  year={2021},
  volume={abs/2110.09193}
}
Unsupervised feature learning often finds low-dimensional embeddings that capture the structure of complex data. For tasks for which prior expert topological knowledge is available, incorporating this into the learned representation may lead to higher quality embeddings. For example, this may help one to embed the data into a given number of clusters, or to accommodate for noise that prevents one from deriving the distribution of the data over the model directly, which can then be learned more… 

References

SHOWING 1-10 OF 55 REFERENCES
A Topology Layer for Machine Learning
TLDR
A differentiable topology layer that computes persistent homology based on level set Filtrations and distance-bases filtrations is presented that can serve as a regularizer directly on data or the weights of machine learning models.
Optimizing persistent homology based functions
TLDR
This work proposes a general framework that allows us to define and compute gradients for persistence-based functions in a very simple way, and provides a simple, explicit and sufficient condition for convergence of stochastic subgradient methods for such functions.
Mining Topological Structure in Graphs through Forest Representations
We consider the problem of inferring simplified topological substructures—which we term backbones—in metric and non-metric graphs. Intuitively, these are subgraphs with ‘few’ nodes, multifurcations,
A Fast and Robust Method for Global Topological Functional Optimization
TLDR
This work introduces a novel backpropagation scheme that is significantly faster, more stable, and produces more robust optima and can be used to produce a stable visualization of dots in a persistence diagram as a distribution over critical, and near-critical, simplices in the data structure.
A roadmap for the computation of persistent homology
TLDR
A friendly introduction to PH is given, the pipeline for the computation of PH is navigated with an eye towards applications, and a range of synthetic and real-world data sets are used to evaluate currently available open-source implementations for the computations of PH.
The Curse Revisited: a Newly Quantified Concept of Meaningful Distances for Learning from High-Dimensional Noisy Data
TLDR
It is derived that under particular conditions, empirical neighborhood relations affected by noise are still likely to be true even when this discrimination to be poor, even when there is a poor relative discrimination between the furthest and closest point.
The Curse Revisited: When are Distances Informative for the Ground Truth in Noisy High-Dimensional Data?
TLDR
It is derived that under particular conditions, empirical neighborhood relations affected by noise are still likely to be truthful even when distance concentration occurs, which means that there is a poor relative discrimination between the furthest and closest neighbors in the data.
Stable topological signatures for metric trees through graph approximations
Graph Approximations to Geodesics on Metric Graphs
TLDR
It is proven that existing conditions under which proximity graphs are ideal representations for a much broader class of spaces, such as metric graphs, cannot be straightforwardly adapted to these spaces.
A comparison of single-cell trajectory inference methods
TLDR
The authors comprehensively benchmark the accuracy, scalability, stability and usability of 45 single-cell trajectory inference methods and develop a set of guidelines to help users select the best method for their dataset.
...
...