Alert button
Picture for Sebastian Damrich

Sebastian Damrich

Alert button

IWR at Heidelberg University

Geometric Autoencoders -- What You See is What You Decode

Jun 30, 2023
Philipp Nazari, Sebastian Damrich, Fred A. Hamprecht

Figure 1 for Geometric Autoencoders -- What You See is What You Decode
Figure 2 for Geometric Autoencoders -- What You See is What You Decode
Figure 3 for Geometric Autoencoders -- What You See is What You Decode
Figure 4 for Geometric Autoencoders -- What You See is What You Decode

Visualization is a crucial step in exploratory data analysis. One possible approach is to train an autoencoder with low-dimensional latent space. Large network depth and width can help unfolding the data. However, such expressive networks can achieve low reconstruction error even when the latent representation is distorted. To avoid such misleading visualizations, we propose first a differential geometric perspective on the decoder, leading to insightful diagnostics for an embedding's distortion, and second a new regularizer mitigating such distortion. Our ``Geometric Autoencoder'' avoids stretching the embedding spuriously, so that the visualization captures the data structure more faithfully. It also flags areas where little distortion could not be achieved, thus guarding against misinterpretation.

* 24 pages, ICML 2023 
Viaarxiv icon

Contrastive learning unifies $t$-SNE and UMAP

Jun 03, 2022
Sebastian Damrich, Jan Niklas Böhm, Fred A. Hamprecht, Dmitry Kobak

Figure 1 for Contrastive learning unifies $t$-SNE and UMAP
Figure 2 for Contrastive learning unifies $t$-SNE and UMAP
Figure 3 for Contrastive learning unifies $t$-SNE and UMAP
Figure 4 for Contrastive learning unifies $t$-SNE and UMAP

Neighbor embedding methods $t$-SNE and UMAP are the de facto standard for visualizing high-dimensional datasets. They appear to use very different loss functions with different motivations, and the exact relationship between them has been unclear. Here we show that UMAP is effectively negative sampling applied to the $t$-SNE loss function. We explain the difference between negative sampling and noise-contrastive estimation (NCE), which has been used to optimize $t$-SNE under the name NCVis. We prove that, unlike NCE, negative sampling learns a scaled data distribution. When applied in the neighbor embedding setting, it yields more compact embeddings with increased attraction, explaining differences in appearance between UMAP and $t$-SNE. Further, we generalize the notion of negative sampling and obtain a spectrum of embeddings, encompassing visualizations similar to $t$-SNE, NCVis, and UMAP. Finally, we explore the connection between representation learning in the SimCLR setting and neighbor embeddings, and show that (i) $t$-SNE can be optimized using the InfoNCE loss and in a parametric setting; (ii) various contrastive losses with only few noise samples can yield competitive performance in the SimCLR setup.

* 29 pages, 13 figures 
Viaarxiv icon

On UMAP's true loss function

Apr 22, 2021
Sebastian Damrich, Fred A. Hamprecht

Figure 1 for On UMAP's true loss function
Figure 2 for On UMAP's true loss function
Figure 3 for On UMAP's true loss function
Figure 4 for On UMAP's true loss function

UMAP has supplanted t-SNE as state-of-the-art for visualizing high-dimensional datasets in many disciplines, but the reason for its success is not well understood. In this work, we investigate UMAP's sampling based optimization scheme in detail. We derive UMAP's effective loss function in closed form and find that it differs from the published one. As a consequence, we show that UMAP does not aim to reproduce its theoretically motivated high-dimensional UMAP similarities. Instead, it tries to reproduce similarities that only encode the shared $k$ nearest neighbor graph, thereby challenging the previous understanding of UMAP's effectiveness. Instead, we claim that the key to UMAP's success is its implicit balancing of attraction and repulsion resulting from negative sampling. This balancing in turn facilitates optimization via gradient descent. We corroborate our theoretical findings on toy and single cell RNA sequencing data.

* 20 pages, 15 figures; minor changes, added run-times and error bars 
Viaarxiv icon

UMAP does not reproduce high-dimensional similarities due to negative sampling

Mar 26, 2021
Sebastian Damrich, Fred A. Hamprecht

Figure 1 for UMAP does not reproduce high-dimensional similarities due to negative sampling
Figure 2 for UMAP does not reproduce high-dimensional similarities due to negative sampling
Figure 3 for UMAP does not reproduce high-dimensional similarities due to negative sampling
Figure 4 for UMAP does not reproduce high-dimensional similarities due to negative sampling

UMAP has supplanted t-SNE as state-of-the-art for visualizing high-dimensional datasets in many disciplines, while the reason for its success is not well understood. In this work, we investigate UMAP's sampling based optimization scheme in detail. We derive UMAP's effective loss function in closed form and find that it differs from the published one. As a consequence, we show that UMAP does not aim to reproduce its theoretically motivated high-dimensional UMAP similarities. Instead, it tries to reproduce similarities that only encode the shared k nearest neighbor graph, thereby challenging the previous understanding of UMAP's effectiveness. Instead, we claim that the key to UMAP's success is its implicit balancing of attraction and repulsion resulting from negative sampling. This balancing in turn facilitates optimization via gradient descent. We corroborate our theoretical findings on toy and single cell RNA sequencing data.

* 18 pages, 14 figures 
Viaarxiv icon

Visualizing hierarchies in scRNA-seq data using a density tree-biased autoencoder

Feb 11, 2021
Quentin Garrido, Sebastian Damrich, Alexander Jäger, Dario Cerletti, Manfred Claassen, Laurent Najman, Fred Hamprecht

Figure 1 for Visualizing hierarchies in scRNA-seq data using a density tree-biased autoencoder
Figure 2 for Visualizing hierarchies in scRNA-seq data using a density tree-biased autoencoder
Figure 3 for Visualizing hierarchies in scRNA-seq data using a density tree-biased autoencoder
Figure 4 for Visualizing hierarchies in scRNA-seq data using a density tree-biased autoencoder

Single cell RNA sequencing (scRNA-seq) data makes studying the development of cells possible at unparalleled resolution. Given that many cellular differentiation processes are hierarchical, their scRNA-seq data is expected to be approximately tree-shaped in gene expression space. Inference and representation of this tree-structure in two dimensions is highly desirable for biological interpretation and exploratory analysis. Our two contributions are an approach for identifying a meaningful tree structure from high-dimensional scRNA-seq data, and a visualization method respecting the tree-structure. We extract the tree structure by means of a density based minimum spanning tree on a vector quantization of the data and show that it captures biological information well. We then introduce DTAE, a tree-biased autoencoder that emphasizes the tree structure of the data in low dimensional space. We compare to other dimension reduction methods and demonstrate the success of our method experimentally. Our implementation relying on PyTorch and Higra is available at github.com/hci-unihd/DTAE.

Viaarxiv icon

MultiStar: Instance Segmentation of Overlapping Objects with Star-Convex Polygons

Nov 26, 2020
Florin C. Walter, Sebastian Damrich, Fred A. Hamprecht

Figure 1 for MultiStar: Instance Segmentation of Overlapping Objects with Star-Convex Polygons
Figure 2 for MultiStar: Instance Segmentation of Overlapping Objects with Star-Convex Polygons
Figure 3 for MultiStar: Instance Segmentation of Overlapping Objects with Star-Convex Polygons
Figure 4 for MultiStar: Instance Segmentation of Overlapping Objects with Star-Convex Polygons

Instance segmentation of overlapping objects in biomedical images remains a largely unsolved problem. We take up this challenge and present MultiStar, an extension to the popular instance segmentation method StarDist. The key novelty of our method is that we identify pixels at which objects overlap and use this information to improve proposal sampling and to avoid suppressing proposals of truly overlapping objects. This allows us to apply the ideas of StarDist to images with overlapping objects, while incurring only a small overhead compared to the established method. MultiStar shows promising results on two datasets and has the advantage of using a simple and easy to train network architecture.

* Submitted to ISBI 2021 
Viaarxiv icon

Probabilistic Watershed: Sampling all spanning forests for seeded segmentation and semi-supervised learning

Nov 06, 2019
Enrique Fita Sanmartin, Sebastian Damrich, Fred A. Hamprecht

Figure 1 for Probabilistic Watershed: Sampling all spanning forests for seeded segmentation and semi-supervised learning
Figure 2 for Probabilistic Watershed: Sampling all spanning forests for seeded segmentation and semi-supervised learning
Figure 3 for Probabilistic Watershed: Sampling all spanning forests for seeded segmentation and semi-supervised learning
Figure 4 for Probabilistic Watershed: Sampling all spanning forests for seeded segmentation and semi-supervised learning

The seeded Watershed algorithm / minimax semi-supervised learning on a graph computes a minimum spanning forest which connects every pixel / unlabeled node to a seed / labeled node. We propose instead to consider all possible spanning forests and calculate, for every node, the probability of sampling a forest connecting a certain seed with that node. We dub this approach "Probabilistic Watershed". Leo Grady (2006) already noted its equivalence to the Random Walker / Harmonic energy minimization. We here give a simpler proof of this equivalence and establish the computational feasibility of the Probabilistic Watershed with Kirchhoff's matrix tree theorem. Furthermore, we show a new connection between the Random Walker probabilities and the triangle inequality of the effective resistance. Finally, we derive a new and intuitive interpretation of the Power Watershed.

* To be published in NeurIPS2019 
Viaarxiv icon