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.

168,657 papers · 148 categories

Trend · papers per month

76151227302 · Jun 202019922001200920172026
48 results for transformer embeddings

Proposes QQE for transforming and embedding data distributions.

problem Transforming and embedding data distributions for better representation or visualization.
method Quantile-Quantile Embedding (QQE) using quantile-quantile plot concept.
result QQE allows for better discrimination of classes in some cases.

Transformers learn to recall with non-orthogonal embeddings in realistic settings.

problem Understanding how transformers store and retrieve knowledge in practical scenarios.
method Analyzing a single-layer transformer with random embeddings trained on a token-retrieval task.
result Explicit formulas for the model's storage capacity reveal a multiplicative dependence on sample size, embedding dimension, and sequence length.

Minimal token perturbations reveal how Transformer models process information.

problem Understanding information propagation in Transformer models for interpretability.
method Study of minimal token perturbations on embedding space.
result Rare tokens cause larger shifts, and input information mixes deeper.

We break down transformer embeddings into interpretable components revealing hidden geometric structures.

problem Understanding the hidden geometry and interpretability of transformer models.
method Decomposed transformer embeddings into position, context, and residual components.
result Pervasive mathematical structure in transformer embeddings, including position and context vectors.

InfiniteWalk connects deep network embeddings to spectral graph theory with a nonlinear transformation.

problem Learning node representations from networks with deep learning methods.
method Study of the DeepWalk objective in the limit as window size goes to infinity, linking to spectral graph embeddings with a nonlinear transformation.
result Simple binary thresholding of the Laplacian pseudoinverse can approximate DeepWalk embeddings.

FREDE efficiently embeds graphs using linear space and guarantees quality.

problem Efficiently embedding graphs with quality guarantees and linear space complexity.
method FREDE combines matrix sketching with a nonlinear transform of PageRank similarities to achieve linear space and quality guarantees.
result FREDE provides column-covariance approximation guarantees that are nearly as good as SVD, even with limited node similarities.

Generative Adversarial Nets (GANs) and Variational Auto-Encoders (VAEs) provide impressive image generations from Gaussian white noise, but the underlying mathematics are not well understood. We compute deep convolutional network generators by inverting a fixed embedding operator. Therefore, they do not require to be o…

2018-05-17abs ↗pdf ↗

When analyzing weighted networks using spectral embedding, a judicious transformation of the edge weights may produce better results. To formalize this idea, we consider the asymptotic behavior of spectral embedding for different edge-weight representations, under a generic low rank model. We measure the quality of dif…

2019-10-12abs ↗pdf ↗

ANT learns sparse embeddings for large vocabularies efficiently.

problem Lack of scalable methods for embedding large vocabularies in neural networks.
method Anchor & Transform (ANT) algorithm that learns a small set of anchor embeddings and a sparse transformation matrix.
result ANT achieves stronger performance with fewer parameters (up to 40x compression) compared to existing methods.

Improved graph embedding through refined linear transformation and community recovery.

problem Identifying meaningful latent communities in graph data.
method Refined graph encoder embedding via linear transformation, self-training, and latent community recovery.
result Improved vertex embedding and better decision boundaries for vertex classification.

Transformers show strengths and weaknesses in complexity analysis.

problem Understanding the strengths and limitations of attention layers in transformers.
method Analysis of representation power through complexity parameters and task-specific constructions.
result Transformers can solve sparse averaging tasks with logarithmic complexity, but triple detection tasks require linear complexity.

Graphs from van der Corput sequence embed into Chamanara surface.

problem Embedding graphs from van der Corput sequence into surfaces.
method Constructed 44-regular graphs from van der Corput sequence and Kronecker sequence, embedded into torus and Chamanara surface.
result Graphs from van der Corput sequence embed into Chamanara surface with one edge removal.

Study geodesic properties of time series data using Wasserstein metric.

problem Modeling nonlinear time series with transport-based metrics.
method Generalized Wasserstein metric and signed cumulative distribution transforms.
result Geodesic properties provide added interpretability and robustness in time series classifiers.

EmDT generates synthetic fraud data to improve detection accuracy.

problem Imbalanced datasets in fraud detection lead to poor performance on rare fraudulent transactions.
method EmDT uses UMAP clustering to identify fraudulent patterns and a Transformer denoising network to generate synthetic data.
result EmDT significantly improves classification performance compared to existing methods.

Tautness of submanifolds in spheres is preserved under Lie sphere transformations.

problem Invariance of tautness under Lie sphere transformations.
method Extending tautness definition to Legendre submanifolds and using Lie sphere transformations to show invariance.
result Tautness is invariant under Lie sphere transformations.

This paper introduces a new task to better understand Transformers in quantitative contexts.

problem Understanding Transformers in high-stakes quantitative and scientific applications.
method Introduces a novel contextual counting task and analyzes it with causal and non-causal Transformer architectures.
result Causal attention is better suited for the contextual counting task, and no positional embeddings lead to the best accuracy.

Transformers encode latent distributions in text, improving performance in out-of-distribution cases.

problem What should embeddings from language models represent?
method Connecting autoregressive prediction to sufficient statistics, identifying three settings.
result Transformers encode latent generating distributions, improving performance.

This work improves KG embeddings by integrating hyperbolic and attention mechanisms.

problem Preserving hierarchical and logical patterns in KGs with low-dimensional embeddings.
method Combines hyperbolic reflections/rotations with attention mechanisms to capture complex relational patterns.
result Improves MRR by up to 6.1% on standard benchmarks and new state-of-the-art results in high dimensions.

This paper clarifies VAE's property through geometric and information-theoretic interpretations.

problem The transparency of VAE model is an underlying issue.
method Quantitative understanding of VAE through differential geometry and information theory.
result VAE can be mapped to an implicit isometric embedding with a scale factor derived from the posterior parameter.

DEN learns diverse tasks to generalize to unseen tasks.

problem Generalization from a diverse set of classification tasks with limited data.
method Three-block architecture: covariate transformation, distribution embedding, and classification.
result DEN outperforms existing methods in various synthetic and real tasks.

BERTopic improves financial text analysis with FinTextSim's contextual embeddings.

problem Analyzing financial text data for insights and predictions.
method Integrates BERTopic with FinTextSim for topic modeling and clustering.
result BERTopic performs better with FinTextSim's embeddings, improving topic clarity and reducing misclassification.

The Joint Optimization of Fidelity and Commensurability (JOFC) manifold matching methodology embeds an omnibus dissimilarity matrix consisting of multiple dissimilarities on the same set of objects. One approach to this embedding optimizes the preservation of fidelity to each individual dissimilarity matrix together wi…

2015-02-11abs ↗pdf ↗

The paper extends optimal transport for linear separability of sheared distributions in supervised learning.

problem Learning on the space of probability measures using shifts and scalings.
method Embedding probability measures into L2L^2 spaces using optimal transport, then applying regular machine learning techniques.
result Sheared distributions can be linearly separated under certain conditions, with bounds on transformations.

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…

2017-12-07abs ↗pdf ↗

A new quantization strategy reduces Transformer model size and inference time.

problem Heavy computation load and memory overhead in Transformer models for mobile devices.
method Mixed precision quantization with varying bits per word in embedding blocks.
result 11.8x smaller model size and 3.5x speed up for on-device NMT.

Word embeddings are commonly obtained as optimizers of a criterion function ff of a text corpus, but assessed on word-task performance using a different evaluation function gg of the test data. We contend that a possible source of disparity in performance on tasks is the incompatibility between classes of transformat…

2019-11-06abs ↗pdf ↗

In this paper, we generalize the famous Hasimoto's transformation by showing that the dynamics of a closed unidimensional vortex filament embedded in a three-dimensional manifold of constant curvature gives rise under Hasimoto's transformation to the non-linear Schrodinger equation. We also give a natural interpretatio…

2012-04-24abs ↗pdf ↗

CoT enhances transformer accuracy on serial tasks by enabling serial computation.

problem Improving accuracy of large language models on inherently serial problems.
method Integrating a chain of thought (CoT) into decoder-only transformers to enable serial computation.
result Constant-depth transformers with CoT can solve problems in AC^0, surpassing TC^0 without CoT.

Although modern recommendation systems can exploit the structure in users' item feedback, most are powerless in the face of new users who provide no structure for them to exploit. In this paper we introduce ImplicitCE, an algorithm for recommending items to new users during their sign-up flow. ImplicitCE works by trans…

2018-09-10abs ↗pdf ↗

This research studies affine invariance in continuous-domain convolutional neural networks.

problem Recognizing patterns and features under affine transformations in continuous domains.
method Introduces a new criterion for assessing affine invariance, embeds images into the affine Lie group, and analyzes convolution over this group.
result Extends the scope of geometrical transformations that deep-learning pipelines can handle.

MPP trains a transformer to predict multiple physical systems, improving accuracy across various tasks.

problem Training models for specific physical systems is inefficient and requires fine-tuning.
method MPP trains a shared transformer on multiple heterogeneous physical systems, projecting fields into a shared embedding space.
result A single MPP-pretrained transformer outperforms task-specific models on all pretraining sub-tasks and downstream tasks.