arXiv research

Method generates visual explanations for similarity models without classification.

RFX-Fuse combines Breiman and Cutler's Random Forest with modern ML capabilities.

SPINEX improves clustering with explainable neighbors, outperforming other methods.

SPINEX improves time series forecasting with explainable neighbors.

SX-GeoTree improves spatially coherent explanations in geospatial regression trees.

BiLRP explains deep similarity models by decomposing scores into feature contributions.

There are some similarities between cohomology of SU(2)-representation varieties of the fundamental group of some link complements and the Khovanov homology of the links. We start here a program to explain a possible source of these similarities. We introduce a symplectic manifold ${\mathscr M}$ with an action of the b…

A new method to explain black box models using Shapley values.

Proposes a method for explaining tabular data using copulas.

This paper explores explaining GBDT2NN predictions by its teacher model, improving distillation performance.

SHAP clustering explains model predictions by grouping similar feature contributions.

Explains differences and similarities of strictly nef and ample vector bundles.

We give explicit formulae for fringe lengths of the Calegari-Walker Ziggurats -- i.e. graphs of extremal rotation numbers associated to positive words in free groups. These formulae reveal (partial) integral projective self-similarity in ziggurat fringes, which are low-dimensional projections of characteristic polyhedr…

Framework explains deep learning candlestick recognition.

Local surrogate explainers vary in objectives, leading to incomparable explanations.

Proposes Coherent Gradients to explain and reduce overfitting in neural networks.

The aim of this paper is to analyze the processes of polarization and agglomeration, to explain the mechanisms and causes of these phenomena in order to identify similarities and differences. As the main implication of this study should be noted that both process pretend to explain the concentration of economic activit…

We describe first integrals of geostrophic equations, which are similar to the enstrophy invariants of the Euler equation for an ideal incompressible fluid. We explain the geometry behind this similarity, give several equivalent definitions of the Poisson structure on the space of smooth densities on a symplectic manif…

TX-Ray analyzes and quantifies model knowledge transfer in NLP.

RELAX provides first attribution-based explanations for representations.

DiSeNE generates interpretable node embeddings without supervision.

Hides the complexity of neural networks, making them more transparent.

This paper explains spectral clustering and its equivalence to PCA, breaking it into fully connected and multi-connected cases.

Study proposes a data-driven CBR system for improved bankruptcy prediction.

Paper proposes a method to make image model explanations robust to distortions.

RATIO improves neural network robustness and explainability.

Study evaluates relevance metrics for similarity-based model explanations.

The Global Vectors for word representation (GloVe), introduced by Jeffrey Pennington et al. is reported to be an efficient and effective method for learning vector representations of words. State-of-the-art performance is also provided by skip-gram with negative-sampling (SGNS) implemented in the word2vec tool. In this…

Unified framework for model explanation methods based on feature removal.

The paper deals with the problem of identifying the internal dependencies and similarities among a large number of random processes. Linear models are considered to describe the relations among the time series and the energy associated to the corresponding modeling error is the criterion adopted to quantify their simil…

Paper explains contrastive learning using cosine similarity and proposes mitigations for batch size effects.

A method to improve clustering explainability using bagging and feature dropout.

Neural networks auto-denoise similar inputs, enabling new statistical analysis.

To study how mental object representations are related to behavior, we estimated sparse, non-negative representations of objects using human behavioral judgments on images representative of 1,854 object categories. These representations predicted a latent similarity structure between objects, which captured most of the…

We introduce a new method to explain Gaussian processes using Shapley values.

We introduce self-dual manifolds and show that they can be used to encode mirror symmetry for affine-Kähler manifolds and for elliptic curves. Their geometric properties, especially the link with special lagrangian fibrations and the existence of a transformation similar to the Fourier-Mukai functor, suggest that this …

The paper studies harmonic map heat flow stability and decay rates.

The paper proposes a new algorithm to select subsets of training data for better accuracy and explainability.

Anchors explain text model decisions by highlighting key words.

To every oriented link $L$, we associate a topologically defined biquandle $\widehat{\mathcal{B}}_{L}$, which we call the topological biquandle of $L$. The construction of $\widehat{\mathcal{B}}_{L}$ is similar to the topological description of the fundamental quandle given by Matveev. We find a presentation of the top…

A conformal map from a Riemann surface to a Euclidean space of dimension greater than or equal to three is explained by using the Clifford algebra, in a similar fashion to quaternionic holomorphic geometry of surfaces in the Euclidean three- or four-space. The Weierstrass representation, the spin transform, the Darboux…

This work improves explanation quality for time series predictions by learning perturbations.

Linear trends in classifier accuracy observed under distribution shift.

Pantypes improve prototypical models by capturing diverse input distributions.

RSM provides insights into deep survival models' decision-making.

A new score function improves explainability and reliability of AI systems.

Predictive modeling applications increasingly use data representing people's behavior, opinions, and interactions. Fine-grained behavior data often has different structure from traditional data, being very high-dimensional and sparse. Models built from these data are quite difficult to interpret, since they contain man…

This paper explores using SSIM for better image generation in generative models.