A5034942394- Topological and Geometric Data Analysis
- Data Visualization and Analytics
- Chronic Obstructive Pulmonary Disease (COPD) Research
- Complex Network Analysis Techniques
- Nutrition, Genetics, and Disease
- Health, Environment, Cognitive Aging
- Morphological variations and asymmetry
- Point processes and geometric inequalities
- Neural Networks and Applications
- 3D Shape Modeling and Analysis
- Advanced Bandit Algorithms Research
- Machine Learning and Algorithms
- Advanced Numerical Analysis Techniques
- Advanced Neuroimaging Techniques and Applications
- T-cell and B-cell Immunology
- Face and Expression Recognition
- Immune Cell Function and Interaction
- Stochastic Gradient Optimization Techniques
- Immunotherapy and Immune Responses
- Advanced Graph Neural Networks
- Reinforcement Learning in Robotics
- Domain Adaptation and Few-Shot Learning
- Markov Chains and Monte Carlo Methods
- Bioinformatics and Genomic Networks
- Artificial Intelligence in Law
Harvard University
2025
Harvard University Press
2022-2024
UC San Diego Health System
2023
California University of Pennsylvania
2023
Princeton University
2017-2022
University of Oxford
2022
Genotype (Germany)
2022
German Rheumatism Research Centre
2011-2020
Leibniz Association
2016-2020
University of Georgia
2018
Abstract At present, it is not clear how memory B lymphocytes are maintained over time, and whether only as circulating cells or also residing in particular tissues. Here we describe distinct populations of isotype-switched (Bsm) murine spleen bone marrow, identified according to individual transcriptional signature cell receptor repertoire. A population marginal zone-like located exclusively the spleen, while a quiescent Bsm found marrow. Three further resident populations, present...
We introduce Forman-Ricci curvature and its corresponding flow as characteristics for complex networks attempting to extend the common approach of node-based network analysis by edge-based characteristics. Following a theoretical introduction mathematical motivation, we apply proposed network-analytic methods static dynamic compare results with established Our work suggests number applications data mining, including denoising clustering experimental data, well extrapolation evolution.
Obsessive-compulsive disorder (OCD) is a common neuropsychiatric disease affecting about 2% of the general population. It characterized by persistent intrusive thoughts and repetitive ritualized behaviors. While gene variations, malfunction cortico-striato-thalamo-cortical (CSTC) circuits, dysregulated synaptic transmission have been implicated in pathogenesis OCD, underlying mechanisms remain largely unknown. Here we show that OCD-like behavior mice caused deficiency SPRED2, protein...
Evaluating gene networks with respect to known biology is a common task but often computationally costly one. Many computational experiments are difficult apply exhaustively in network analysis due run-times. To permit high-throughput of networks, we have implemented set very efficient tools calculate functional properties based on guilt-by-association methods. ( xtending ' uilt-by- ssociation' by egree) allows be evaluated hundreds or thousands sets. The methods predict novel members...
Abstract The notion of curvature on graphs has recently gained traction in the networks community, with Ollivier–Ricci (ORC) particular being used for several tasks network analysis, such as community detection. In this work, we choose a different approach and study augmentations discretization Ricci proposed by Forman (AFRC). We empirically theoretically investigate its relation to ORC un-augmented Forman–Ricci curvature. particular, provide evidence that AFRC frequently gives sufficient...
Abstract A cornerstone of machine learning is the identification and exploitation structure in high‐dimensional data. While classical approaches assume that data lies a Euclidean space, geometric methods are designed for non‐Euclidean data, including graphs, strings, matrices, or characterized by symmetries inherent underlying system. In this article, we review uncovering leveraging how an understanding geometry can lead to development more effective algorithms with provable guarantees.
Higher-order information is crucial for relational learning in many domains where relationships extend beyond pairwise interactions. Hypergraphs provide a natural framework modeling such relationships, which has motivated recent extensions of graph neural net- work architectures to hypergraphs. However, comparisons between hypergraph and standard graph-level models remain limited. In this work, we systematically evaluate selection hypergraph-level architectures, determine their effectiveness...
We present a viable geometric solution for the detection of dynamic effects in complex networks. Building on Forman’s discretization classical notion Ricci curvature, we introduce novel method to characterize different types real-world networks with an emphasis peer-to-peer study Ricci-flow network-theoretic setting and analytic tool characterizing effects. The formalism suggests computational change identification fast evolving network regions yields insights into topological properties...
In T lymphocytes, expression of miR-148a is induced by T-bet and Twist1, specific for pro-inflammatory Th1 cells. these cells, inhibits the pro-apoptotic protein Bim promotes their survival. Here we use sequence-specific cholesterol-modified oligonucleotides against (antagomir-148a) selective elimination cells in vivo. murine model transfer colitis, antagomir-148a treatment reduced number colon colitic mice 50% inhibited 71% remaining Expression colonic was increased. Antagomir-148a-mediated...
Abstract We study projection-free methods for constrained Riemannian optimization. In particular, we propose a Frank-Wolfe ( RFW ) method that handles constraints directly, in contrast to prior rely on (potentially costly) projections. analyze non-asymptotic convergence rates of an optimum geodesically convex problems, and critical point nonconvex objectives. also present practical setting under which can attain linear rate. As concrete example, specialize the manifold positive definite...
Traditionally, network analysis is based on local properties of vertices, like their degree or clustering, and statistical behaviour across the in question. This article develops an approach which different two respects: we investigate edge-based properties, define global characteristics networks directly. More concretely, start with Forman's notion Ricci curvature a graph, more generally, polyhedral complex. will allow us to pass from graph as representing complex for instance by filling...
Abstract Background To date, most studies involving high-throughput analyses of sputum in asthma and COPD have focused on identifying transcriptomic signatures disease. No whole-genome methylation analysis cells has been performed yet. In this context, the highly variable cellular composition potential to confound molecular analyses. Methods Whole-genome transcription (Agilent Human 4 × 44 k array) (Illumina 450 BeadChip) were samples 9 asthmatics, 10 healthy subjects. RNA integrity was...
Recently, there has been a surge of interest in representation learning hyperbolic spaces, driven by their ability to represent hierarchical data with significantly fewer dimensions than standard Euclidean spaces. However, the viability and benefits spaces for downstream machine tasks have received less attention. In this paper, we present, our knowledge, first theoretical guarantees classifier rather space. Specifically, consider problem large-margin possessing structure. We provide an...
Proinflammatory type 1 T helper (Th1) cells are enriched in inflamed tissue and contribute to the maintenance of chronic inflammation rheumatic diseases. Here we show that microRNA- (miR-) 31 is upregulated murine Th1 with a history repeated reactivation memory Th isolated from synovial fluid patients joint disease. Knock-down miR-31 resulted upregulation genes associated cytoskeletal rearrangement motility induced expression target involved cell activation, chemokine receptor-...
A key challenge in Machine Learning (ML) is the identification of geometric structure high-dimensional data. Most algorithms assume that data lives a vector space; however, many applications involve non-Euclidean data, such as graphs, strings and matrices, or whose determined by symmetries underlying system. Here, we discuss methods for identifying how leveraging geometry can give rise to efficient ML with provable guarantees.
The manifold hypothesis presumes that high-dimensional data lies on or near a low-dimensional manifold. While the utility of encoding geometric structure has been demonstrated empirically, rigorous analysis its impact learnability neural networks is largely missing. Several recent results have established hardness for learning feedforward and equivariant under i.i.d. Gaussian uniform Boolean distributions. In this paper, we investigate hypothesis. We ask which minimal assumptions curvature...
Graph Machine Learning often involves the clustering of nodes based on similarity structure encoded in graph's topology and nodes' attributes. On homophilous graphs, integration pooling layers has been shown to enhance performance Neural Networks by accounting for inherent multi-scale structure. Here, similar are grouped together coarsen graph reduce input size subsequent deeper architectures. In both settings, underlying approach can be implemented via operators, which rely classical tools...
The human brain forms functional networks on all spatial scales. Modern fMRI scanners allow to resolve data in high resolutions, allowing study large-scale that relate cognitive processes. analysis of such a cornerstone experimental neuroscience. Due the immense size and complexity underlying sets, efficient evaluation visualization remain challenge for analysis. In this study, we combine recent advances neuroscience applied mathematics perform mathematical characterization complex...
Abstract Summary Evaluating gene networks with respect to known biology is a common task but often computationally costly one. Many computational experiments are difficult apply exhaustively in network analysis due run-times. To permit high-throughput of networks, we have implemented set very efficient tools calculate functional properties based on guilt-by-association methods. EGAD ( E xtending ‘ G uilt-by- A ssociation’ by D egree) allows be evaluated hundreds or thousands sets. The...