Vision models excel in image classification but struggle to generalize to unseen data, such as classifying images from unseen domains or discovering novel categories. In this paper, we explore the relationship between logical reasoning and deep learning generalization in visual classification. A logical regularization termed L-Reg is derived which bridges a logical analysis framework to image classification. Our work reveals that L-Reg reduces the complexity of the model in terms of the feature distribution and classifier weights. Specifically, we unveil the interpretability brought by L-Reg, as it enables the model to extract the salient features, such as faces to persons, for classification. Theoretical analysis and experiments demonstrate that L-Reg enhances generalization across various scenarios, including multi-domain generalization and generalized category discovery. In complex real-world scenarios where images span unknown classes and unseen domains, L-Reg consistently improves generalization, highlighting its practical efficacy.

Decision-making processes have increasingly come to rely on sophisticated machine learning tools, raising concerns about the fairness of their predictions with respect to any sensitive groups. The widespread use of commercial black-box machine learning models necessitates careful consideration of their legal and ethical implications on consumers. In situations where users have access to these "black-box" models, a key question emerges: how can we mitigate or eliminate the influence of sensitive attributes, such as race or gender? We propose towerDebias (tDB), a novel approach designed to reduce the influence of sensitive variables in predictions made by black-box models. Using the Tower Property from probability theory, tDB aims to improve prediction fairness during the post-processing stage in a manner amenable to the Fairness-Utility Tradeoff. This method is highly flexible, requiring no prior knowledge of the original model's internal structure, and can be extended to a range of different applications. We provide a formal improvement theorem for tDB and demonstrate its effectiveness in both regression and classification tasks, underscoring its impact on the fairness-utility tradeoff.

Neural responses encode information that is useful for a variety of downstream tasks. A common approach to understand these systems is to build regression models or ``decoders'' that reconstruct features of the stimulus from neural responses. Popular neural network similarity measures like centered kernel alignment (CKA), canonical correlation analysis (CCA), and Procrustes shape distance, do not explicitly leverage this perspective and instead highlight geometric invariances to orthogonal or affine transformations when comparing representations. Here, we show that many of these measures can, in fact, be equivalently motivated from a decoding perspective. Specifically, measures like CKA and CCA quantify the average alignment between optimal linear readouts across a distribution of decoding tasks. We also show that the Procrustes shape distance upper bounds the distance between optimal linear readouts and that the converse holds for representations with low participation ratio. Overall, our work demonstrates a tight link between the geometry of neural representations and the ability to linearly decode information. This perspective suggests new ways of measuring similarity between neural systems and also provides novel, unifying interpretations of existing measures.

Integral projection models (IPMs) are widely used to study population growth and the dynamics of demographic structure (e.g. age and size distributions) within a population.These models use data on individuals' growth, survival, and reproduction to predict changes in the population from one time point to the next and use these in turn to ask about long-term growth rates, the sensitivity of that growth rate to environmental factors, and aspects of the long term population such as how much reproduction concentrates in a few individuals; these quantities are not directly measurable from data and must be inferred from the model. Building IPMs requires us to develop models for individual fates over the next time step -- Did they survive? How much did they grow or shrink? Did they Reproduce? -- conditional on their initial state as well as on environmental covariates in a manner that accounts for the unobservable quantities that are are ultimately interested in estimating.Targeted maximum likelihood estimation (TMLE) methods are particularly well-suited to a framework in which we are largely interested in the consequences of models. These build machine learning-based models that estimate the probability distribution of the data we observe and define a target of inference as a function of these. The initial estimate for the distribution is then modified by tilting in the direction of the efficient influence function to both de-bias the parameter estimate and provide more accurate inference. In this paper, we employ TMLE to develop robust and efficient estimators for properties derived from a fitted IPM. Mathematically, we derive the efficient influence function and formulate the paths for the least favorable sub-models. Empirically, we conduct extensive simulations using real data from both long term studies of Idaho steppe plant communities and experimental Rotifer populations.

We offer theoretical and empirical insights into the impact of exogenous randomness on the effectiveness of random forests with tree-building rules independent of training data. We formally introduce the concept of exogenous randomness and identify two types of commonly existing randomness: Type I from feature subsampling, and Type II from tie-breaking in tree-building processes. We develop non-asymptotic expansions for the mean squared error (MSE) for both individual trees and forests and establish sufficient and necessary conditions for their consistency. In the special example of the linear regression model with independent features, our MSE expansions are more explicit, providing more understanding of the random forests' mechanisms. It also allows us to derive an upper bound on the MSE with explicit consistency rates for trees and forests. Guided by our theoretical findings, we conduct simulations to further explore how exogenous randomness enhances random forest performance. Our findings unveil that feature subsampling reduces both the bias and variance of random forests compared to individual trees, serving as an adaptive mechanism to balance bias and variance. Furthermore, our results reveal an intriguing phenomenon: the presence of noise features can act as a "blessing" in enhancing the performance of random forests thanks to feature subsampling.

Causal structures play a central role in world models that flexibly adapt to changes in the environment. While recent works motivate the benefits of discovering local causal graphs for dynamics modelling, in this work we demonstrate that accurately capturing these relationships in complex settings remains challenging for the current state-of-the-art. To remedy this shortcoming, we postulate that sparsity is a critical ingredient for the discovery of such local causal structures. To this end we present the SPARse TrANsformer World model (SPARTAN), a Transformer-based world model that learns local causal structures between entities in a scene. By applying sparsity regularisation on the attention pattern between object-factored tokens, SPARTAN identifies sparse local causal models that accurately predict future object states. Furthermore, we extend our model to capture sparse interventions with unknown targets on the dynamics of the environment. This results in a highly interpretable world model that can efficiently adapt to changes. Empirically, we evaluate SPARTAN against the current state-of-the-art in object-centric world models on observation-based environments and demonstrate that our model can learn accurate local causal graphs and achieve significantly improved few-shot adaptation to changes in the dynamics of the environment as well as robustness against removing irrelevant distractors.

Transformer-based large language models (LLMs) have displayed remarkable creative prowess and emergence capabilities. Existing empirical studies have revealed a strong connection between these LLMs' impressive emergence abilities and their in-context learning (ICL) capacity, allowing them to solve new tasks using only task-specific prompts without further fine-tuning. On the other hand, existing empirical and theoretical studies also show that there is a linear regularity of the multi-concept encoded semantic representation behind transformer-based LLMs. However, existing theoretical work fail to build up an understanding of the connection between this regularity and the innovative power of ICL. Additionally, prior work often focuses on simplified, unrealistic scenarios involving linear transformers or unrealistic loss functions, and they achieve only linear or sub-linear convergence rates. In contrast, this work provides a fine-grained mathematical analysis to show how transformers leverage the multi-concept semantics of words to enable powerful ICL and excellent out-of-distribution ICL abilities, offering insights into how transformers innovate solutions for certain unseen tasks encoded with multiple cross-concept semantics. Inspired by empirical studies on the linear latent geometry of LLMs, the analysis is based on a concept-based low-noise sparse coding prompt model. Leveraging advanced techniques, this work showcases the exponential 0-1 loss convergence over the highly non-convex training dynamics, which pioneeringly incorporates the challenges of softmax self-attention, ReLU-activated MLPs, and cross-entropy loss. Empirical simulations corroborate the theoretical findings.

We consider the problem of sampling from a $d$-dimensional log-concave distribution $\pi(\theta) \propto \exp(-f(\theta))$ for $L$-Lipschitz $f$, constrained to a convex body with an efficiently computable self-concordant barrier function, contained in a ball of radius $R$ with a $w$-warm start. We propose a \emph{robust} sampling framework that computes spectral approximations to the Hessian of the barrier functions in each iteration. We prove that for polytopes that are described by $n$ hyperplanes, sampling with the Lee-Sidford barrier function mixes within $\widetilde O((d^2+dL^2R^2)\log(w/\delta))$ steps with a per step cost of $\widetilde O(nd^{\omega-1})$, where $\omega\approx 2.37$ is the fast matrix multiplication exponent. Compared to the prior work of Mangoubi and Vishnoi, our approach gives faster mixing time as we are able to design a generalized soft-threshold Dikin walk beyond log-barrier. We further extend our result to show how to sample from a $d$-dimensional spectrahedron, the constrained set of a semidefinite program, specified by the set $\{x\in \mathbb{R}^d: \sum_{i=1}^d x_i A_i \succeq C \}$ where $A_1,\ldots,A_d, C$ are $n\times n$ real symmetric matrices. We design a walk that mixes in $\widetilde O((nd+dL^2R^2)\log(w/\delta))$ steps with a per iteration cost of $\widetilde O(n^\omega+n^2d^{3\omega-5})$. We improve the mixing time bound of prior best Dikin walk due to Narayanan and Rakhlin that mixes in $\widetilde O((n^2d^3+n^2dL^2R^2)\log(w/\delta))$ steps.

We demonstrate that Gini coefficients can be used as unified metrics to evaluate many-versus-many (all-to-all) similarity in vector spaces. Our analysis of various image datasets shows that images with the highest Gini coefficients tend to be the most similar to one another, while images with the lowest Gini coefficients are the least similar. We also show that this relationship holds true for vectorized text embeddings from various corpuses, highlighting the consistency of our method and its broad applicability across different types of data. Additionally, we demonstrate that selecting machine learning training samples that closely match the distribution of the testing dataset is far more important than ensuring data diversity. Selection of exemplary and iconic training samples with higher Gini coefficients leads to significantly better model performance compared to simply having a diverse training set with lower Gini coefficients. Thus, Gini coefficients can serve as effective criteria for selecting machine learning training samples, with our selection method outperforming random sampling methods in very sparse information settings.

Central to achieving the energy transition, heating systems provide essential space heating and hot water in residential and industrial environments. A major challenge lies in effectively profiling large clusters of buildings to improve demand estimation and enable efficient Demand Response (DR) schemes. This paper addresses this challenge by introducing an unsupervised machine learning framework for clustering residential heating load profiles, focusing on natural gas space heating and hot water preparation boilers. The profiles are analyzed across five dimensions: boiler usage, heating demand, weather conditions, building characteristics, and user behavior. We apply three distance metrics: Euclidean Distance (ED), Dynamic Time Warping (DTW), and Derivative Dynamic Time Warping (DDTW), and evaluate their performance using established clustering indices. The proposed method is assessed considering 29 residential buildings in Greece equipped with smart meters throughout a calendar heating season (i.e., 210 days). Results indicate that DTW is the most suitable metric, uncovering strong correlations between boiler usage, heat demand, and temperature, while ED highlights broader interrelations across dimensions and DDTW proves less effective, resulting in weaker clusters. These findings offer key insights into heating load behavior, establishing a solid foundation for developing more targeted and effective DR programs.