arXiv research

A new method optimizes slicing directions for SW distances to improve high-dimensional probability measure comparison.

Magnetic manifold HMC improves sampling on constrained manifolds.

The kernel embedding algorithm is an important component for adapting kernel methods to large datasets. Since the algorithm consumes a major computation cost in the testing phase, we propose a novel teacher-learner framework of learning computation-efficient kernel embeddings from specific data. In the framework, the h…

We construct embedded Willmore tori with small area constraint in Riemannian three-manifolds under some curvature condition used to prevent Möbius degeneration. The construction relies on a Lyapunov-Schmidt reduction; to this aim we establish new geometric expansions of exponentiated small symmetric Clifford tori and a…

Visual rendering of graphs is a key task in the mapping of complex network data. Although most graph drawing algorithms emphasize aesthetic appeal, certain applications such as travel-time maps place more importance on visualization of structural network properties. The present paper advocates two graph embedding appro…

Detects drifts in data for classification tasks using constrained embeddings.

Dimensionality reduction is a topic of recent interest. In this paper, we present the classification constrained dimensionality reduction (CCDR) algorithm to account for label information. The algorithm can account for multiple classes as well as the semi-supervised setting. We present an out-of-sample expressions for …

Unified Lagrangian-based methods for nonsmooth nonconvex optimization.

PoET-BiN reduces power consumption in neural networks on embedded devices.

TASO optimizes CNN models for memory-constrained devices.

Data aggregation improves HAC for resource-constrained systems.

Visual rendering of graphs is a key task in the mapping of complex network data. Although most graph drawing algorithms emphasize aesthetic appeal, certain applications such as travel-time maps place more importance on visualization of structural network properties. The present paper advocates a graph embedding approac…

CLEANN detects and mitigates neural network Trojans without labeled data.

Delaunay tori minimize Willmore energy under isoperimetric constraints.

Paper proposes IIQ for compressing embedding vectors.

ALF reduces network parameters and operations by 70% and 61%, respectively, on embedded hardware.

Graph-based RKD improves knowledge distillation for resource-constrained systems.

Graph embedding aims at learning a vector-based representation of vertices that incorporates the structure of the graph. This representation then enables inference of graph properties. Existing graph embedding techniques, however, do not scale well to large graphs. We therefore propose a framework for parallel computat…

Suppose $M$ is a complete, embedded minimal surface in $\mathbb{R}^3$ with an infinite number of ends, finite genus and compact boundary. We prove that the simple limit ends of $M$ have properly embedded representatives with compact boundary, genus zero and with constrained geometry. We use this result to show that if …

ClauseLens uses reinforcement learning to price reinsurance treaties transparently and auditably.

Efficient method for constrained optimization under partial observations with provable convergence.

Algorithm optimizes constrained reinforcement learning with dual variables.

Enhanced GNN with expanded attention window and partially random embeddings.

Physics-constrained deep learning predicts geophysical dynamics with boundedness.

This work optimizes quantization of linear models to reduce memory usage.

Dual pairs constructed for volume preserving diffeomorphisms using symplectic geometry.

Study compares constrained and decoupled moduli spaces of manifolds with particles and discs.

Paper introduces privacy-preserving few-shot learning for images.

Convolutional neural networks have recently achieved significant breakthroughs in various image classification tasks. However, they are computationally expensive,which can make their feasible mplementation on embedded and low-power devices difficult. In this paper convolutional neural network binarization is implemente…

Compressing word embeddings is important for deploying NLP models in memory-constrained settings. However, understanding what makes compressed embeddings perform well on downstream tasks is challenging---existing measures of compression quality often fail to distinguish between embeddings that perform well and those th…

Embeddings are one of the fundamental building blocks for data analysis tasks. Embeddings are already essential tools for large language models and image analysis, and their use is being extended to many other research domains. The generation of these distributed representations is often a data- and computation-expensi…

Well-quasi-orders proved on embedded planar graphs.

Efficient neural networks for resource-constrained systems.

Parametric UMAP learns a mapping from data to embeddings.

For every $g\in\mathbb{N}_0$ and $ε>0$, we construct a smooth genus $g$ surface embedded into the unit ball with area $8π$ and Willmore energy smaller than $8π+ ε$. From this we deduce that a minimising sequence for Willmore's energy in the class of genus $g$ surfaces embedded in the unit ball with area $8π$ converges …

Near isometric orthogonal embeddings to lower dimensions are a fundamental tool in data science and machine learning. In this paper, we present the construction of such embeddings that minimizes the maximum distortion for a given set of points. We formulate the problem as a non convex constrained optimization problem. …

Graph embedding is a popular algorithmic approach for creating vector representations for individual vertices in networks. Training these algorithms at scale is important for creating embeddings that can be used for classification, ranking, recommendation and other common applications in industry. While industrial syst…

Metric learning methods for dimensionality reduction in combination with k-Nearest Neighbors (kNN) have been extensively deployed in many classification, data embedding, and information retrieval applications. However, most of these approaches involve pairwise training data comparisons, and thus have quadratic computat…

Simple non-convex methods outperform others in ordinal embedding.

This is the second of a series of two papers where we construct embedded Willmore tori with small area constraint in Riemannian three-manifolds. In both papers the construction relies on a Lyapunov-Schmidt reduction, the difficulty being the Möbius degeneration of the tori. In the first paper the construction was perfo…

Study of embeddings avoiding certain tangent patterns using polynomial spaces.

Enhanced PC$^2$ improves surrogate modeling for high-dimensional problems.

MEDAL converts manifold embeddings into models for rigorous validation.

The recent advances in deep neural networks (DNNs) make them attractive for embedded systems. However, it can take a long time for DNNs to make an inference on resource-constrained computing devices. Model compression techniques can address the computation issue of deep inference on embedded devices. This technique is …

The paper uncovers symmetries in large language models through layer-peeled optimization.

We demonstrate that graphs embedded on surfaces are a powerful and practical tool to generate, characterize and simulate networks with a broad range of properties. Remarkably, the study of topologically embedded graphs is non-restrictive because any network can be embedded on a surface with sufficiently high genus. The…

Adaptive regularization prevents overfitting in large-scale sparse feature models.

For every $\;b>1\;$ fixed, we explicitly construct $1$-dimensional families of embedded constrained Willmore tori parametrized by their conformal class $\;(a,b)$\; with $\; a \sim_b 0^+\;$ deforming the homogenous torus \;$f^b$ of conformal class \;$(0,b).$ The variational vector field at $f^b$ is hereby given by a non…