Method converts speech with attention and context preservation.
problem Voice conversion with improved stability and efficiency.
method Sequence-to-Sequence learning with attention and context preservation.
result Synthesized speech quality comparable to advanced methods.
Paper proposes dp-VAE for preserving spatial context in gene expression data.
problem Inaccessibility of spatial context in single-cell gene expression data.
method Generic representation learning and transfer learning framework with a distance-preserving regularizer.
result dp-VAE effectively reconstructs and imputes spatial context from gene expression data.
Paper discusses privacy issues in IoT and proposes a lightweight neural network approach.
problem Privacy concerns in IoT due to extensive data collection and processing.
method Developed a privacy-preserving inference approach for IoT objects and a deep neural network in the cloud.
result Satisfactory performance of the proposed approach on the MNIST dataset.
In this note we consider the relationship between the dressing action and the holonomy representation in the context of constant mean curvature surfaces. We characterize dressing elements that preserve the topology of a surface and discuss dressing by simple factors as a means of adding bubbles to a class of non finite…
Transformers preserve support and can approximate any continuous map.
problem Understanding the mathematical properties of transformers.
method Characterizing maps between measures that can be represented as transformers and proving their properties.
result Transformers preserve support and have uniformly continuous Fréchet derivatives.
We prove an existence result for local and global G-structure preserving affine immersions between affine manifolds. Several examples are discussed in the context of Riemannian and semi-Riemannian geometry, including the case of isometric immersions into Lie groups endowed with a left-invariant metric, and the case of …
The paper studies quasimorphisms on density-preserving diffeomorphisms of the Möbius band.
problem Exploring quasimorphisms on groups of diffeomorphisms of non-orientable manifolds.
method Investigates the group of density-preserving diffeomorphisms on the Möbius band and shows the existence of unbounded quasimorphisms.
result The group of density-preserving diffeomorphisms on the Möbius band admits countably many unbounded quasimorphisms.
MCNet improves uncertainty calibration in online advertising by modeling complex relations and balancing performance.
problem Lack of effective calibration for complex relations and context features in online advertising.
method Introduces MCNet with MCF, order-preserving, and field-balance regularizers.
result Superior performance in generating well-calibrated probability predictions on public and industrial datasets.
In this paper we provide a systematic discussion of how to incorporate orientation preserving symmetries into the treatment of Willmore surfaces via the loop group method. In this context we first develop a general treatment of Willmore surfaces admitting orientation preserving symmetries, and then show how to induce f…
Language models learn from training data and can leak private information.
problem Language models lack context understanding and can expose private data.
method Discussing the limitations of current privacy protection methods for language models.
result Existing privacy protection methods are insufficient for language models.
IVFS simplifies feature selection for high-dimensional data preservation.
problem Maintaining structure and pairwise distances in high-dimensional data.
method IVFS framework based on persistent diagrams from computational topology.
result IVFS well preserves pairwise distances and topological patterns of full data.
Proposes differentially private normalizing flows for privacy-preserving density estimation.
problem Privacy concerns in density estimation models when individuals are directly associated with the training data.
method Uses normalizing flow models with explicit differential privacy guarantees.
result Substantially outperforms previous state-of-the-art approaches in privacy-preserving density estimation.
Study preserves planar and graphical properties of curves under elastic flow.
problem Maintaining planar and graphical properties of non-compact curves under elastic flow.
method Extended recent work on adapted elastic energy to derive thresholds for planar and graphical embeddedness.
result Derived new Li--Yau type inequality for complete planar curves.
HCFContext predicts mobile context using collaborative filtering and homomorphic encryption.
problem Accurate mobile context determination for enterprise policies.
method Proposes HPContext and HCFContext models using sequential history and collaborative filtering, with privacy-preserving homomorphic encryption.
result HCFContext enhances context prediction by leveraging related users' observations.
Study quantifies context dependency in image classification and segmentation models.
problem Understanding how much context affects model predictions in image classification and segmentation.
method Developed a method to quantify and control model sensitivity to visual context by removing selected objects from images.
result Discovered that certain objects (e.g., 'sidewalk') rely heavily on the presence of other objects (e.g., 'cars') in the context.
This study extends a result on quasi-isometry of hyperbolic groups to relatively hyperbolic groups.
problem Classifying groups up to quasi-isometry, focusing on relatively hyperbolic groups.
method Defining quasiconformal maps and showing their equivalence to coarsely cusp-preserving quasi-isometries between Bowditch boundaries.
result Quasiconformal maps between Bowditch boundaries of relatively hyperbolic groups are equivalent to coarsely cusp-preserving quasi-isometries.
New method distinguishes feature relevance in non-linear contexts.
problem Finding relevant features with preserved redundancies.
method Random forest models and statistical methods.
result Distinguishes strong from weak feature relevance in non-linear problems.
In this paper we develop a general conceptual approach to the problem of existence of action-angle variables for dynamical systems, which establishes and uses the fundamental conservation property of associated torus actions: anything which is preserved by the system is also preserved by the associated torus actions. T…
Fine-tuning harms in-context learning, but restricting updates to the value matrix improves zero-shot performance.
problem Fine-tuning harms in-context learning, reducing zero-shot performance on unseen tasks.
method Theoretical analysis of linear attention models, identifying conditions for degraded few-shot performance.
result Restricting updates to the value matrix improves zero-shot performance while preserving in-context learning.
Study preserves symplectic structure in forced discrete mechanical systems.
problem Preserving symplectic structure in forced discrete mechanical systems.
method Analyzes a specific type of forced discrete mechanical system (Q,Ld,fd), preserving a symplectic structure on QimesQ. result The preserved symplectic structure can be seen as Marsden-Weinstein reduction of the canonical symplectic structure.
CADE learns dual node representations for better generalization.
problem Transductive graph embeddings cannot generalize to unseen nodes or across different graphs.
method CADE combines real-time neighborhoods with neighbor-attentioned representation, preserving known node memory.
result CADE outperforms state-of-the-art methods in generalization and context-awareness.
Metrics of exceptional holonomy are vacuum solutions to the Einstein equation. In this paper we describe manifolds with holonomy contained in Spin(7) preserved by a three-torus symmetry in terms of tri-symplectic geometry of four-manifolds. These complement examples that have appeared in the context of domain wall prob…
In this paper we prove that an isometry between orbit spaces of two proper isometric actions is smooth if it preserves the codimension of the orbits or if the orbit spaces have no boundary. In other words, we generalize Myers-Steenrod's theorem for orbit spaces. These results are proved in the more general context of s…
Geometric deformations preserve post-Lie algebra structure in regularity structures.
problem Deriving geometric deformations of post-Lie algebras.
method Extending geometrical notions of torsion and curvature, deriving compatibility conditions.
result Derives a pre-Lie structure for regularity structures, isomorphic to a post-Lie algebra.
Study on Goeritz equivalence in genus 2 Heegaard splitting of S3.
problem Understanding Goeritz equivalence of curves in genus 2 Heegaard splitting of S3. method Introduce Goeritz equivalence of curves, present algebraic obstructions, and provide examples.
result Algebraic obstructions to Goeritz equivalence of simple closed curves are computed and demonstrated.
A new framework uses directed information to efficiently select context chunks.
problem Efficiently selecting relevant context chunks for query understanding.
method Directed Information γ-covering framework, formulated as a γ-cover problem, with a greedy algorithm for context selection. result The γ-covering algorithm provides clear advantages in hard-decision regimes like context compression and single-slot prompt selection. Unified theory of measure-preserving diffusions on manifolds.
problem Deriving a complete recipe for measure-preserving diffusions on manifolds.
method Developed a geometric theory that unifies and generalizes previous constructions, relying on intrinsic geometry of the target measure.
result The completeness result is a direct consequence of manifold topology and target measure geometry.
Causal analysis reveals regional discrepancies in TOPCAT trial results.
problem Inconclusive results in TOPCAT trial for heart failure treatment.
method Causal discovery methods with domain knowledge integration.
result Significant causal effects shown for some subgroups globally.
Paper proposes a privacy-preserving knockoff inference method.
problem Ensuring privacy in model-X knockoff inference.
method Differential privacy framework for knockoff inference.
result Guaranteed FDR control with privacy protection.
Improving in-context learning for latent space Bayesian optimization by adapting pretraining on molecular latent space.
problem Improving in-context learning for latent space Bayesian optimization.
method Adapting pretraining on molecular latent space.
result Achieving strong performance on held-out molecular optimization benchmarks.
Survey of privacy-preserving distributed deep learning methods.
problem Protecting confidential patterns in data during distributed deep learning.
method Comparison of federated learning, split learning, large batch SGD, and privacy-preserving techniques.
result Trade-offs between computational resources, data leakage, and communication efficiency.
We propose a new notion called \emph{infinity-harmonic maps}between Riemannain manifolds. These are natural generalizations of the well known notion of infinity harmonic functions and are also the limiting case of p% -harmonic maps as p→∞. Infinity harmoncity appears in many familiar contexts. For example,…
A new method learns Hamiltonian functions from noisy data.
problem Learning Hamiltonian functions from noisy observations.
method Structure-preserving kernel ridge regression method.
result The method yields excellent numerical performances.
LLMs compress financial texts, but distort decision-making.
problem LLMs compress financial texts, altering decision-making.
method Analyzed two diagnostic patterns: decontextualization and model dependency. Proposed Agentic Context Compression.
result LLM-compressed financial texts alter decision-making.
We study combinations of risk measures under no restrictive assumption on the set of alternatives. We develop and discuss results regarding the preservation of properties and acceptance sets for the combinations of risk measures. One of the main results is the representation of resulting risk measures from the properti…
Identifying context-specific entity networks from aggregated data is an important task, arising often in bioinformatics and neuroimaging. Computationally, this task can be formulated as jointly estimating multiple different, but related, sparse Undirected Graphical Models (UGM) from aggregated samples across several co…
Combines machine learning and convex limiting for accurate subgrid flux modeling in shallow-water equations.
problem Accurate subgrid flux modeling in shallow-water equations.
method Machine learning and flux limiting for property-preserving subgrid scale modeling.
result The proposed method produces meaningful closures even in untrained scenarios.
Genus 2 mutation is the process of cutting a 3-manifold along an embedded closed genus 2 surface, twisting by the hyper-elliptic involution, and gluing back. This paper compares genus 2 mutation with the better-known Conway mutation in the context of knots in the 3-sphere. Despite the fact that any Conway mutation can …
When a complex semisimple group G acts holomorphically on a Kähler manifold (X,ω) such that a maximal compact subgroup K⊂G preserves the symplectic form ω, a basic result of symplectic geometry says that the corresponding categorical quotient X/G can be identified with quotient of the zero-set of the m…
We develop a comprehensive mathematical framework for polynomial jump-diffusions in a semimartingale context, which nest affine jump-diffusions and have broad applications in finance. We show that the polynomial property is preserved under polynomial transformations and Lévy time change. We present a generic method for…
The success of graph embeddings or node representation learning in a variety of downstream tasks, such as node classification, link prediction, and recommendation systems, has led to their popularity in recent years. Representation learning algorithms aim to preserve local and global network structure by identifying no…
Around 1960, R. Palais and J. Cerf proved a fundamental result relating spaces of diffeomorphisms and imbeddings of manifolds: If V is a submanifold of M, then the map from Diff(M) to Imb(V,M) that takes f to its restriction to V is locally trivial. We extend this and related results into the context of fibered manifol…
We extend the coherent state transform (CST) of Hall to the context of the moduli spaces of semistable holomorphic vector bundles with fixed determinant over elliptic curves. We show that by applying the CST to appropriate distributions, we obtain the space of level k, rank n and genus one non-abelian theta functions w…
InstaGAN tackles image-to-image translation for images with multiple instances and significant shape changes.
problem Challenging cases, especially images with multiple target instances and significant shape changes.
method Instance-aware GAN (InstaGAN) that incorporates instance information and context preserving loss.
result Improves multi-instance transfiguration while maintaining permutation invariance of instances.
We present a Donaldson-Witten type field theory in eight dimensions on manifolds with Spin(7) holonomy. We prove that the stress tensor is BRST exact for metric variations preserving the holonomy and we give the invariants for this class of variations. In six and seven dimensions we propose similar theories on Calabi…
Attention temperature improves robustness of ICL in high-dimensional settings.
problem ICL robustness failure under distribution shift in high dimensions.
method Analyzed a Transformer with approximate softmax attention, derived a closed-form error expression, and showed optimal temperature minimizes error.
result Optimal attention temperature minimizes ICL generalization error under distribution shift.
Paper studies optimal federated learning for nonparametric regression with privacy constraints.
problem Federated learning for nonparametric regression with heterogeneous differential privacy constraints.
method Proposes distributed privacy-preserving estimators and investigates their risk properties.
result Establishes matching minimax lower bounds for global and pointwise estimation.
The paper proposes differentially private sliced inverse regression algorithms for high-dimensional data.
problem Privacy concerns in high-dimensional data analysis.
method Differentially private sliced inverse regression algorithms designed for privacy preservation.
result Achieves minimax lower bounds up to logarithmic factors.