Research
On-device research index

arXiv research

A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.

168,786 papers · 148 categories

Trend · papers per month

6.3%12.5%18.8%25.0% · Oct 199319922001200920172026
48 results for Image-text classification

Unsupervised methods have proven effective for discriminative tasks in a single-modality scenario. In this paper, we present a multimodal framework for learning sparse representations that can capture semantic correlation between modalities. The framework can model relationships at a higher level by forcing the shared …

2015-11-19abs ↗pdf ↗

Global pooling, such as max- or sum-pooling, is one of the key ingredients in deep neural networks used for processing images, texts, graphs and other types of structured data. Based on the recent DeepSets architecture proposed by Zaheer et al. (NIPS 2017), we introduce a Set Aggregation Network (SAN) as an alternative…

2018-10-03abs ↗pdf ↗

The paper analyzes and proposes an algorithm for multi-modal nonlinear embeddings with theoretical performance bounds.

problem Generalizability of multi-modal nonlinear embeddings to unseen data.
method Theoretical analysis and a multi-modal nonlinear representation learning algorithm motivated by performance bounds.
result The proposed algorithm yields promising performance in multi-modal image classification and cross-modal image-text retrieval applications.

In this paper we study output coding for multi-label prediction. For a multi-label output coding to be discriminative, it is important that codewords for different label vectors are significantly different from each other. In the meantime, unlike in traditional coding theory, codewords in output coding are to be predic…

2012-06-27abs ↗pdf ↗

Proposes MR-SNE for multimodal data visualization.

problem Visualizing data from multiple domains with relations across them.
method Extends t-SNE to compute augmented relations and jointly embed them in a low-dimensional space.
result Demonstrates promising performance in visualizing Flickr and Animal with Attributes 2 datasets.

New method improves neural network performance by focusing on steep function regions.

problem Improving neural network performance by focusing on steep function regions.
method Variance Based Samples Weighting (VBSW) using labels local variance to weight training points.
result Significantly increases the performances of neural networks for various tasks.

Adaptive weighting schemes enhance time-series data augmentation for financial and UCR datasets.

problem Limited size of time-series datasets hinders model performance.
method Two adaptive weighting schemes for automatic data augmentation.
result Improves annualized returns by over 50% on financial dataset and outperforms state-of-the-art on half of UCR datasets.

Unified principle LZN unifies generative modeling, representation learning, and classification.

problem Disjoint solutions for generative modeling, representation learning, and classification.
method LZN creates a shared latent space for all tasks, using encoders and decoders for each data type.
result LZN improves FID on CIFAR10 and outperforms existing methods in representation learning and joint generation/classification.

ZegOT uses optimal transport to zero-shot segment images with text prompts.

problem Zero-shot semantic segmentation with limited image-text alignment knowledge.
method ZegOT uses optimal transport to match multiple text prompts with frozen image embeddings.
result ZegOT achieves state-of-the-art performance in zero-shot semantic segmentation.

Contrastive learning estimates transition kernels for continuous-time stochastic processes.

problem Estimating transition kernels for continuous-time stochastic processes without labeled data.
method Contrastive learning applied to strong-mixing continuous-time stochastic processes.
result Contrastive learning can estimate transition kernels for small-to-mid-range intervals in the diffusion case.

Traditional classifiers can generate high-quality images comparable to generative models.

problem Separation between classifiers and generators in neural networks.
method Optimizing input gradients to produce images, using mask-based stochastic reconstruction, progressive-resolution technique, and distance metric loss.
result Traditional classifiers can generate high-fidelity images of 256imes imes256 resolution on ImageNet.

A new framework for consistent segmentation evaluation reduces operating losses.

problem Inconsistent thresholding-based segmentation methods lead to suboptimal solutions.
method Developed a consistent ranking-based framework (RankDice/RankIoU) using Bayes rules and Dice-/IoU-calibration.
result The proposed framework is Dice-/IoU-calibrated and provides excess risk bounds and convergence rates.

Estimates CATEs for structured treatments using a new decomposition method.

problem Estimating conditional average treatment effects for complex data types.
method Generalized Robinson decomposition, isolating causal estimand, arbitrary model plugging, quasi-oracle convergence guarantee.
result Demonstrates superior performance in CATE estimation compared to prior work.

Symile learns joint representations across multiple modalities, outperforming pairwise CLIP.

problem Pairwise contrastive learning fails to capture joint information between multiple modalities.
method Symile uses a flexible, architecture-agnostic objective to learn modality-specific representations by deriving a lower bound on total correlation.
result Symile outperforms pairwise CLIP on cross-modal classification and retrieval across various datasets.

Since its introduction, unsupervised representation learning has attracted a lot of attention from the research community, as it is demonstrated to be highly effective and easy-to-apply in tasks such as dimension reduction, clustering, visualization, information retrieval, and semi-supervised learning. In this work, we…

2018-11-05abs ↗pdf ↗

Classifies and constructs intertwining differential operators between line and vector bundles over real projective space.

problem Classifying and constructing intertwining differential operators between line and vector bundles over real projective space.
method F-method for classification and construction of intertwining differential operators.
result Generalizes a classical result of Bol for SL(2,R)SL(2,\mathbb{R}) and classifies intertwining operators for SL(n,R)SL(n,\mathbb{R}).

The paper classifies and constructs differential symmetry breaking operators from a line bundle to a vector bundle over real projective spaces.

problem Classifying and constructing differential symmetry breaking operators.
method Utilizing factorization identities and branching laws of generalized Verma modules.
result Differential symmetry breaking operators from a line bundle to a vector bundle over real projective spaces are classified and constructed.

We propose JECL, a method for clustering image-caption pairs by training parallel encoders with regularized clustering and alignment objectives, simultaneously learning both representations and cluster assignments. These image-caption pairs arise frequently in high-value applications where structured training data is e…

2019-01-04abs ↗pdf ↗

Traditional clustering methods often perform clustering with low-level indiscriminative representations and ignore relationships between patterns, resulting in slight achievements in the era of deep learning. To handle this problem, we develop Deep Discriminative Clustering (DDC) that models the clustering task by inve…

2019-05-05abs ↗pdf ↗

This chapter is dedicated to the assessment and performance estimation of machine learning (ML) algorithms, a topic that is equally important to the construction of these algorithms, in particular in the context of cyberphysical security design. The literature is full of nonparametric methods to estimate a statistic fr…

2019-07-30abs ↗pdf ↗

While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations fro…

2018-07-10abs ↗pdf ↗

New method extracts factors of variation from data without much supervision.

problem Disentangling complex sensory inputs into simple factors of variation without much supervision.
method Develops a new approach for disentanglement under structural assumptions, reducing the need for auxiliary information.
result Disentanglement is possible even when auxiliary information does not ensure conditional independence, with less auxiliary information required.

CLIP learns joint image-text representations for zero-shot learning.

problem Understanding and improving zero-shot transfer performance in CLIP.
method Formal study of transferrable representation learning and analysis of zero-shot transfer performance.
result Proposes a new CLIP-type approach that outperforms existing methods.

We study how well one can recover sparse principal components of a data matrix using a sketch formed from a few of its elements. We show that for a wide class of optimization problems, if the sketch is close (in the spectral norm) to the original data matrix, then one can recover a near optimal solution to the optimiza…

2015-03-12abs ↗pdf ↗

A simple strategy prevents negative transfer in transfer learning.

problem Negative transfer in transfer learning where source representations harm target performance.
method Residual feature integration with a trainable target-side encoder.
result The method provably prevents negative transfer with theoretical guarantees.

The paper discusses quantifying realism in generated images.

problem Designing functions to reliably distinguish realistic data from unrealistic data.
method Drawing on insights from algorithmic information theory, the paper introduces the notion of a universal critic.
result A good generative model alone is insufficient to solve the problem of realism quantification.

Heterogeneous domain adaptation (HDA) aims to facilitate the learning task in a target domain by borrowing knowledge from a heterogeneous source domain. In this paper, we propose a Soft Transfer Network (STN), which jointly learns a domain-shared classifier and a domain-invariant subspace in an end-to-end manner, for a…

2019-08-28abs ↗pdf ↗

The paper tackles calibrating long-term behaviors with multiple styles using programmatic style-consistency.

problem Generating long-term sequential behaviors with multiple styles simultaneously.
method Leverage programmatic labeling functions to specify controllable styles and derive style-consistency as a learning objective.
result Learned policies can be calibrated for up to 1024 distinct style combinations.

This research explores how different discrete diffusion kernels affect graph generation quality.

problem The impact of different discrete diffusion kernels on graph generation quality.
method Developed a family of discrete diffusion kernels that converge to different Bernoulli priors.
result The quality of generated graphs is sensitive to the prior used, challenging previous intuitions.

Transformer is a popularly used neural network architecture, especially for language understanding. We introduce an extended and unified architecture that can be used for tasks involving a variety of modalities like image, text, videos, etc. We propose a spatio-temporal cache mechanism that enables learning spatial dim…

2019-07-17abs ↗pdf ↗

Improved guarantees for sparse random embeddings with explicit bounds and empirical superiority.

problem Improving the explicitness and sharpness of guarantees for sparse random embeddings.
method Explicit bounds, tighter estimates for quadratic chaos, extreme properties of sparse linear forms, and improved bounds for sums of independent random variables.
result Significantly outperforms prior works on various real-world datasets.

Study shows how to reduce data needed for learning under geometric constraints.

problem Learning high-dimensional data with geometric priors.
method Spherical harmonic decompositions and kernel methods for invariance and geometric stability.
result Improvements in sample complexity by leveraging group invariance, with asymptotic behavior depending on spectral properties.

COBRA reduces modality gap in cross-modal tasks.

problem Joint embedding spaces fail to sufficiently reduce modality gap in multi-modal tasks.
method COBRA trains image and text modalities in a joint fashion using Contrastive Predictive Coding and Noise Contrastive Estimation.
result COBRA significantly reduces the modality gap and generates robust joint-embedding space.