arXiv subjects

José M. F. Moura

Publications and source records attributed to José M. F. Moura.

At least 19 recordsLinked to original sources

arXiv2026

Full-Covariance Smoothing of Bayesian Neural Networks for Online Adaptation

A neural network's layers can be treated as time steps of a state-space model, turning Bayesian training into a smoothing problem: a forward pass propagates Gaussian moments through the network, and a backward Rauch--Tung--Striebel pass updates the weight posteriors in closed form. Such methods learn from each observation in a single pass, in an uncertainty-aware manner, and without gradient-based iterations or replay, which makes them well suited for online adaptation and data-efficient learning. Existing smoothing-based methods, however, are restricted to diagonal covariances across activations, discarding correlations between neurons. We overcome this limitation via a cross-covariance identity that enables full-covariance propagation through a network's nonlinear activations. We derive a one-step-per-layer smoother that approximates as Gaussian only each layer's affine output, and that applies both to deterministic systems with noisy observations and to stochastic systems described by output statistics. We demonstrate this method in non-stationary classification, online dynamics learning, and policy adaptation of a vision-language-action model, and find that it is generally more accurate than other smoothing-based methods.

cs.LG↗

arXiv2026

Black-Box Inference of LLM Architectural Properties with Restrictive API Access

In practice, most commercial LLM providers do not publicly release details of underlying LLM architectures. However, prior work has shown that given limited API access to an LLM (namely, top-$k$ logits and/or a logit bias function), one can recover certain architectural details of an LLM, such as the hidden dimension of the feed-forward network. Perhaps in response to these results, most commercial LLM providers have restricted their APIs to expose only the single logit for each decoded token, and they no longer give users the ability to bias logits. We show that even under current restrictive APIs, several architectural parameters are still recoverable. We present NightVision, an attack that uses restrictive black-box API access to estimate the hidden dimension, depth, and parameter count of an LLM. Algorithmically, NightVision relies on a novel common set prompting technique in which multiple prompts expose log probabilities for the same set of output tokens; a spectral analysis of these results is used to infer hidden dimension. NightVision additionally uses end-to-end time to first token (TTFT) measurements and the estimated hidden dimension to estimate depth and parameter count. We empirically evaluate NightVision on 32 open-source LLMs, recovering hidden dimension to within 23% average relative error across all models (9% on MoE models), and depth and parameter count to within 53% for models exceeding three billion parameters. We run extensive ablations to demonstrate how these accuracies scale with token budget and model properties. Overall, our results suggest that current LLM APIs are not sufficiently restricted to fully obfuscate the architectural details of their underlying models.

cs.LG↗

arXiv2026

Representation Matters in Randomized Smoothing for Audio Classification

Randomized smoothing (RS) certifies robustness in the vector space where Gaussian noise is added. In audio classification, this space is often not uniquely defined as standard pipelines normalize, range-control, and transform waveforms into log-mel or other spectral features. We show that direct RS is therefore under-specified unless the certified object and preprocessing policy are explicit. On two audio benchmarks, keyword spotting and environmental-sound classification, we study waveform, feature-space, and post-processed smoothing. Our diagnostics show why representation-aware reporting is necessary: at the same smoothing level $σ=0.0025$, the two datasets share the same median raw radius $.007996$, but different waveform energies yield different SNR-equivalent scales ($83.98$ vs. $90.97$ dB); log-mel smoothing gives higher positive-radius certified accuracy on environmental sounds ($68.42\%$ vs. $65.53\%$), certifying more examples with nonzero radius but over features rather than waveforms; and clipping or peak normalization changes the effective perturbation norm by roughly $230$--$351\times$. We therefore recommend that audio RS studies choose and report the task-specific certified object and perturbation model, including the perturbation location, gain policy, raw radius, and any post-noise geometry changes.

eess.AS↗

arXiv2026

RRISE: Robust Radius Inference via a Surrogate Estimator

Randomized smoothing (RS) uses a smoothed classifier to provide architecture-agnostic certificates of $\ell_2$ classification robustness, but its dependence on per-input Monte Carlo (MC) sampling undermines its use in real-time systems. We argue that this cost is structural rather than fundamental, such that it can be significantly reduced by sharing information across the deployment stream. We introduce RRISE, an RS framework that compresses certification into a single forward pass through a learned surrogate. RRISE trains the surrogate against precomputed MC class-count targets via a soft-label cross-entropy loss and converts surrogate predictions into provably conservative certified radii through a one-time conformal calibration step. The resulting certificate is deployment-verifiable: whenever the calibrated radius is positive, the surrogate's prediction provably matches the smoothed classifier's and the smoothed classifier is constant on a ball of that radius around the input. Across image classification benchmarks, RRISE matches fixed-budget MC certified accuracy within $0.84$ percentage points while replacing up to $10^4$ noisy base-model evaluations per query with a single surrogate forward pass, recouping MC training cost after $\approx 10^5$ deployment queries. On CIFAR-100 and Tiny ImageNet, where the only prior offline-surrogate method collapses, RRISE achieves $1.23$ to $1.91\times$ higher certified accuracy, establishing efficient randomized smoothing as a practical path to certified robustness in repeated-deployment settings.

cs.LG↗

arXiv2026

Preorder Construct on Simple Undirected Graphs

We construct a novel preorder on the set of nodes of a simple undirected graph. We prove that the preorder (induced by the topology of the graph) is preserved, e.g., by the logistic dynamical system (both in discrete and continuous time). Moreover, the underlying equivalence relation of the preorder corresponds to the coarsest equitable partition (CEP). This will further imply that the logistic dynamical system on a graph preserves its coarsest equitable partition. The results provide a nontrivial invariant set for the logistic and the like dynamical systems, as we show. We note that our construct provides a functional characterization for the CEP as an alternative to the pure set theoretical iterated degree sequences characterization. The construct and results presented might have independent interest for analysis on graphs or qualitative analysis of dynamical systems over networks.

math.CO↗

arXiv2026

The Powers of Precision: Structure-Informed Detection in Complex Systems -- From Customer Churn to Seizure Onset

Emergent phenomena -- onset of epileptic seizures, sudden customer churn, or pandemic outbreaks -- often arise from hidden causal interactions in complex systems. We propose a machine learning method for their early detection that addresses a core challenge: unveiling and harnessing a system's latent causal structure despite the data-generating process being unknown and partially observed. The method learns an optimal feature representation from a one-parameter family of estimators -- powers of the empirical covariance or precision matrix -- offering a principled way to tune in to the underlying structure driving the emergence of critical events. A supervised learning module then classifies the learned representation. We prove structural consistency of the family and demonstrate the empirical soundness of our approach on seizure detection and churn prediction, attaining competitive results in both. Beyond prediction, and toward explainability, we ascertain that the optimal covariance power exhibits evidence of good identifiability while capturing structural signatures, thus reconciling predictive performance with interpretable statistical structure.

cs.LG↗

arXiv2026

GLUE: Gradient-free Learning to Unify Experts

In many deployed systems (multilingual ASR, cross-hospital imaging, region-specific perception), multiple pretrained specialist models coexist. Yet, new target domains often require domain expansion: a generalized model that performs well beyond any single specialist's domain. Given a new target domain, existing methods obtain a single strong initialization prior for the model parameters by blending expert models to initialize a target model. However, heuristic blending -- using mixing coefficients based on data size or proxy metrics -- often yields lower target-domain test accuracy, and learning these coefficients on the target domain's loss function typically requires computationally-expensive full backpropagation through a neural network. We propose GLUE, Gradient-free Learning to Unify Experts, which initializes the target model as a convex combination of fixed experts and learns the mixture coefficients of this combination via gradient-free two-point SPSA (simultaneous perturbation stochastic approximation) updates, requiring only two forward passes per step. Across experiments on three datasets and three network architectures, GLUE produces model parameter priors that can be fine-tuned to outperform baselines. GLUE improves test accuracy by up to 8.5% over data-size weighting and by up to 9.1% over proxy-metric selection. GLUE either outperforms backpropagation-based full-gradient mixing or matches its performance within 1.4%.

cs.LG↗

arXiv2025

GradNetOT: Learning Optimal Transport Maps with GradNets

Monotone gradient functions play a central role in solving the Monge formulation of the optimal transport (OT) problem, which arises in modern applications ranging from fluid dynamics to robot swarm control. When the transport cost is the squared Euclidean distance, Brenier's theorem guarantees that the unique optimal transport map satisfies a Monge-Ampère equation and is the gradient of a convex function. In [arXiv:2301.10862] [arXiv:2404.07361], we proposed Monotone Gradient Networks (mGradNets), neural networks that directly parameterize the space of monotone gradient maps. In this work, we leverage mGradNets to directly learn the optimal transport mapping by minimizing a training loss function defined using the Monge-Ampère equation. We empirically show that the structural bias of mGradNets facilitates the learning of optimal transport maps across both image morphing tasks and high-dimensional OT problems.

cs.LG↗

arXiv2025

Kalman Bayesian Transformer

Sequential fine-tuning of transformers is useful when new data arrive sequentially, especially with shifting distributions. Unlike batch learning, sequential learning demands that training be stabilized despite a small amount of data by balancing new information and previously learned knowledge in the pre-trained models. This challenge is further complicated when training is to be completed in latency-critical environments and learning must additionally quantify and be mediated by uncertainty. Motivated by these challenges, we propose a novel method that frames sequential fine-tuning as a posterior inference problem within a Bayesian framework. Our approach integrates closed-form moment propagation of random variables, Kalman Bayesian Neural Networks, and Taylor approximations of the moments of softmax functions. By explicitly accounting for pre-trained models as priors and adaptively balancing them against new information based on quantified uncertainty, our method achieves robust and data-efficient sequential learning. The effectiveness of our method is demonstrated through numerical simulations involving sequential adaptation of a decision transformer to tasks characterized by distribution shifts and limited memory resources.

cs.LG↗

arXiv2025

Inferring the Graph Structure of Images for Graph Neural Networks

Image datasets such as MNIST are a key benchmark for testing Graph Neural Network (GNN) architectures. The images are traditionally represented as a grid graph with each node representing a pixel and edges connecting neighboring pixels (vertically and horizontally). The graph signal is the values (intensities) of each pixel in the image. The graphs are commonly used as input to graph neural networks (e.g., Graph Convolutional Neural Networks (Graph CNNs) [1, 2], Graph Attention Networks (GAT) [3], GatedGCN [4]) to classify the images. In this work, we improve the accuracy of downstream graph neural network tasks by finding alternative graphs to the grid graph and superpixel methods to represent the dataset images, following the approach in [5, 6]. We find row correlation, column correlation, and product graphs for each image in MNIST and Fashion-MNIST using correlations between the pixel values building on the method in [5, 6]. Experiments show that using these different graph representations and features as input into downstream GNN models improves the accuracy over using the traditional grid graph and superpixel methods in the literature.

eess.IV↗

arXiv2025

FedBaF: Federated Learning Aggregation Biased by a Foundation Model

Foundation models are now a major focus of leading technology organizations due to their ability to generalize across diverse tasks. Existing approaches for adapting foundation models to new applications often rely on Federated Learning (FL) and disclose the foundation model weights to clients when using it to initialize the global model. While these methods ensure client data privacy, they compromise model and information security. In this paper, we introduce Federated Learning Aggregation Biased by a Foundation Model (FedBaF), a novel method for dynamically integrating pre-trained foundation model weights during the FL aggregation phase. Unlike conventional methods, FedBaF preserves the confidentiality of the foundation model while still leveraging its power to train more accurate models, especially in non-IID and adversarial scenarios. Our comprehensive experiments use Pre-ResNet and foundation models like Vision Transformer to demonstrate that FedBaF not only matches, but often surpasses the test accuracy of traditional weight initialization methods by up to 11.4% in IID and up to 15.8% in non-IID settings. Additionally, FedBaF applied to a Transformer-based language model significantly reduced perplexity by up to 39.2%.

cs.LG↗

arXiv2025

Peer-to-Peer Learning Dynamics of Wide Neural Networks

Peer-to-peer learning is an increasingly popular framework that enables beyond-5G distributed edge devices to collaboratively train deep neural networks in a privacy-preserving manner without the aid of a central server. Neural network training algorithms for emerging environments, e.g., smart cities, have many design considerations that are difficult to tune in deployment settings -- such as neural network architectures and hyperparameters. This presents a critical need for characterizing the training dynamics of distributed optimization algorithms used to train highly nonconvex neural networks in peer-to-peer learning environments. In this work, we provide an explicit characterization of the learning dynamics of wide neural networks trained using popular distributed gradient descent (DGD) algorithms. Our results leverage both recent advancements in neural tangent kernel (NTK) theory and extensive previous work on distributed learning and consensus. We validate our analytical results by accurately predicting the parameter and error dynamics of wide neural networks trained for classification tasks.

cs.LG↗

arXiv2025

BRIDGES: Bridging Graph Modality and Large Language Models within EDA Tasks

While many EDA tasks already involve graph-based data, existing LLMs in EDA primarily either represent graphs as sequential text, or simply ignore graph-structured data that might be beneficial like dataflow graphs of RTL code. Recent studies have found that LLM performance suffers when graphs are represented as sequential text, and using additional graph information significantly boosts performance. To address these challenges, we introduce BRIDGES, a framework designed to incorporate graph modality into LLMs for EDA tasks. BRIDGES integrates an automated data generation workflow, a solution that combines graph modality with LLM, and a comprehensive evaluation suite. First, we establish an LLM-driven workflow to generate RTL and netlist-level data, converting them into dataflow and netlist graphs with function descriptions. This workflow yields a large-scale dataset comprising over 500,000 graph instances and more than 1.5 billion tokens. Second, we propose a lightweight cross-modal projector that encodes graph representations into text-compatible prompts, enabling LLMs to effectively utilize graph data without architectural modifications. Experimental results demonstrate 2x to 10x improvements across multiple tasks compared to text-only baselines, including accuracy in design retrieval, type prediction and perplexity in function description, with negligible computational overhead (<1% model weights increase and <30% additional runtime overhead). Even without additional LLM finetuning, our results outperform text-only by a large margin. We plan to release BRIDGES, including the dataset, models, and training flow.

cs.LG↗

arXiv2025

ReLU Networks as Random Functions: Their Distribution in Probability Space

This paper presents a novel framework for understanding trained ReLU networks as random, affine functions, where the randomness is induced by the distribution over the inputs. By characterizing the probability distribution of the network's activation patterns, we derive the discrete probability distribution over the affine functions realizable by the network. We extend this analysis to describe the probability distribution of the network's outputs. Our approach provides explicit, numerically tractable expressions for these distributions in terms of Gaussian orthant probabilities. Additionally, we develop approximation techniques to identify the support of affine functions a trained ReLU network can realize for a given distribution of inputs. Our work provides a framework for understanding the behavior and performance of ReLU networks corresponding to stochastic inputs, paving the way for more interpretable and reliable models.

cs.LG↗

arXiv2025

GSA-Forecaster: Forecasting Graph-Based Time-Dependent Data with Graph Sequence Attention

Forecasting graph-based, time-dependent data has broad practical applications but presents challenges. Effective models must capture both spatial and temporal dependencies in the data, while also incorporating auxiliary information to enhance prediction accuracy. In this paper, we identify limitations in current state-of-the-art models regarding temporal dependency handling. To overcome this, we introduce GSA-Forecaster, a new deep learning model designed for forecasting in graph-based, time-dependent contexts. GSA-Forecaster utilizes graph sequence attention, a new attention mechanism proposed in this paper, to effectively manage temporal dependencies. GSA-Forecaster integrates the data's graph structure directly into its architecture, addressing spatial dependencies. Additionally, it incorporates auxiliary information to refine its predictions further. We validate its performance using real-world graph-based, time-dependent datasets, where it demonstrates superior effectiveness compared to existing state-of-the-art models.

cs.LG↗

arXiv2025

Gradient Networks

Directly parameterizing and learning gradients of functions has widespread significance, with specific applications in inverse problems, generative modeling, and optimal transport. This paper introduces gradient networks (GradNets): novel neural network architectures that parameterize gradients of various function classes. GradNets exhibit specialized architectural constraints that ensure correspondence to gradient functions. We provide a comprehensive GradNet design framework that includes methods for transforming GradNets into monotone gradient networks (mGradNets), which are guaranteed to represent gradients of convex functions. Our results establish that our proposed GradNet (and mGradNet) universally approximate the gradients of (convex) functions. Furthermore, these networks can be customized to correspond to specific spaces of potential functions, including transformed sums of (convex) ridge functions. Our analysis leads to two distinct GradNet architectures, GradNet-C and GradNet-M, and we describe the corresponding monotone versions, mGradNet-C and mGradNet-M. Our empirical results demonstrate that these architectures provide efficient parameterizations and outperform existing methods by up to 15 dB in gradient field tasks and by up to 11 dB in Hamiltonian dynamics learning tasks.

cs.LG↗

arXiv2024

Peer-to-Peer Deep Learning for Beyond-5G IoT

We present P2PL, a practical multi-device peer-to-peer deep learning algorithm that, unlike the federated learning paradigm, does not require coordination from edge servers or the cloud. This makes P2PL well-suited for the sheer scale of beyond-5G computing environments like smart cities that otherwise create range, latency, bandwidth, and single point of failure issues for federated approaches. P2PL introduces max norm synchronization to catalyze training, retains on-device deep model training to preserve privacy, and leverages local inter-device communication to implement distributed consensus. Each device iteratively alternates between two phases: 1) on-device learning and 2) peer-to-peer cooperation where they combine model parameters with nearby devices. We empirically show that all participating devices achieve the same test performance attained by federated and centralized training -- even with 100 devices and relaxed singly stochastic consensus weights. We extend these experimental results to settings with diverse network topologies, sparse and intermittent communication, and non-IID data distributions.

cs.LG↗

arXiv2024

An Analytic Solution to Covariance Propagation in Neural Networks

Uncertainty quantification of neural networks is critical to measuring the reliability and robustness of deep learning systems. However, this often involves costly or inaccurate sampling methods and approximations. This paper presents a sample-free moment propagation technique that propagates mean vectors and covariance matrices across a network to accurately characterize the input-output distributions of neural networks. A key enabler of our technique is an analytic solution for the covariance of random variables passed through nonlinear activation functions, such as Heaviside, ReLU, and GELU. The wide applicability and merits of the proposed technique are shown in experiments analyzing the input-output distributions of trained neural networks and training Bayesian neural networks.

cs.LG↗