arXiv subjects

Simone Rossi

Publications and source records attributed to Simone Rossi.

At least 19 recordsLinked to original sources

arXiv2026

DrivingVoxels: Compositional Sparse Voxel Rasterization for Dynamic Driving Scene Reconstruction

Reconstructing dynamic urban scenes remains challenging due to the unbounded nature of driving environments and the presence of multiple dynamic objects. Currently, potentially faster sparse voxel methods are mainly designed for static scenarios. On the other hand, dynamic approaches based on 3D Gaussian Splatting, despite their high-fidelity, are often time-consuming for driving scenarios and exhibit uncontrollable memory growth in large scenes. To address these limitations, we present DrivingVoxels, a compositional sparse voxel rendering framework for dynamic driving scenes. Our method jointly rasterizes sparse voxels from multiple independent octrees within a single rendering pass. Each rigid dynamic object is represented by an octree defined in its local coordinate frame, while a separate static octree models the stationary background. DrivingVoxels adopts a fully explicit, neural-free representation together with a LiDAR-guided structural initialization that efficiently captures scene geometry. We evaluate our framework on the PandaSet benchmark, demonstrating that DrivingVoxels performs on par on perceptual metrics and better on structural metrics for NVS and reconstruction while requiring shorter training times than previous 3DGS-base methods to an efficient optimization workflow anchored by a strong LiDAR prior.

cs.CV↗

arXiv2026

A Bayesian Perspective on the Role of Epistemic Uncertainty for Delayed Generalization in In-Context Learning

In-context learning enables transformers to adapt to new tasks from a few examples at inference time, while grokking highlights that this generalization can emerge abruptly only after prolonged training. We study task generalization and grokking in in-context learning using a Bayesian perspective, asking what enables the delayed transition from memorization to generalization. Concretely, we consider modular arithmetic tasks in which a transformer must infer a latent linear function solely from in-context examples and analyze how predictive uncertainty evolves during training. We combine approximate Bayesian techniques to estimate the posterior distribution and we study how uncertainty behaves across training and under changes in task diversity, context length, and context noise. We find that epistemic uncertainty collapses sharply when the model groks, making uncertainty a practical label-free diagnostic of generalization in transformers. Additionally, we provide theoretical support with a simplified Bayesian linear model, showing that asymptotically both delayed generalization and uncertainty peaks arise from the same underlying spectral mechanism, which links grokking time to uncertainty dynamics.

stat.ML↗

arXiv2025

The Initialization Determines Whether In-Context Learning Is Gradient Descent

In-context learning (ICL) in large language models (LLMs) is a striking phenomenon, yet its underlying mechanisms remain only partially understood. Previous work connects linear self-attention (LSA) to gradient descent (GD), this connection has primarily been established under simplified conditions with zero-mean Gaussian priors and zero initialization for GD. However, subsequent studies have challenged this simplified view by highlighting its overly restrictive assumptions, demonstrating instead that under conditions such as multi-layer or nonlinear attention, self-attention performs optimization-like inference, akin to but distinct from GD. We investigate how multi-head LSA approximates GD under more realistic conditions specifically when incorporating non-zero Gaussian prior means in linear regression formulations of ICL. We first extend multi-head LSA embedding matrix by introducing an initial estimation of the query, referred to as the initial guess. We prove an upper bound on the number of heads needed for ICL linear regression setup. Our experiments confirm this result and further observe that a performance gap between one-step GD and multi-head LSA persists. To address this gap, we introduce yq-LSA, a simple generalization of single-head LSA with a trainable initial guess yq. We theoretically establish the capabilities of yq-LSA and provide experimental validation on linear regression tasks, thereby extending the theory that bridges ICL and GD. Finally, inspired by our findings in the case of linear regression, we consider widespread LLMs augmented with initial guess capabilities, and show that their performance is improved on a semantic similarity task.

cs.LG↗

arXiv2025

Scaling Laws for Uncertainty in Deep Learning

Deep learning has recently revealed the existence of scaling laws, demonstrating that model performance follows predictable trends based on dataset and model sizes. Inspired by these findings and fascinating phenomena emerging in the over-parameterized regime, we examine a parallel direction: do similar scaling laws govern predictive uncertainties in deep learning? In identifiable parametric models, such scaling laws can be derived in a straightforward manner by treating model parameters in a Bayesian way. In this case, for example, we obtain $O(1/N)$ contraction rates for epistemic uncertainty with respect to the number of data $N$. However, in over-parameterized models, these guarantees do not hold, leading to largely unexplored behaviors. In this work, we empirically show the existence of scaling laws associated with various measures of predictive uncertainty with respect to dataset and model sizes. Through experiments on vision and language tasks, we observe such scaling laws for in- and out-of-distribution predictive uncertainty estimated through popular approximate Bayesian inference and ensemble methods. Besides the elegance of scaling laws and the practical utility of extrapolating uncertainties to larger data or models, this work provides strong evidence to dispel recurring skepticism against Bayesian approaches: "In many applications of deep learning we have so much data available: what do we need Bayes for?". Our findings show that "so much data" is typically not enough to make epistemic uncertainty negligible.

stat.ML↗

arXiv2025

From predictions to confidence intervals: an empirical study of conformal prediction methods for in-context learning

Transformers have become a standard architecture in machine learning, demonstrating strong in-context learning (ICL) abilities that allow them to learn from the prompt at inference time. However, uncertainty quantification for ICL remains an open challenge, particularly in noisy regression tasks. This paper investigates whether ICL can be leveraged for distribution-free uncertainty estimation, proposing a method based on conformal prediction to construct prediction intervals with guaranteed coverage. While traditional conformal methods are computationally expensive due to repeated model fitting, we exploit ICL to efficiently generate confidence intervals in a single forward pass. Our empirical analysis compares this approach against ridge regression-based conformal methods, showing that conformal prediction with in-context learning (CP with ICL) achieves robust and scalable uncertainty estimates. Additionally, we evaluate its performance under distribution shifts and establish scaling laws to guide model training. These findings bridge ICL and conformal prediction, providing a theoretically grounded and new framework for uncertainty quantification in transformer-based models.

stat.ML↗

arXiv2025

Think2SQL: Reinforce LLM Reasoning Capabilities for Text2SQL

Large Language Models (LLMs) can translate natural language into SQL, but small models struggle with multi-table and complex queries in Zero-Shot Learning (ZSL) settings. While Supervised Fine-Tuning (SFT) helps, it falls short for harder cases. To address this, we study how different reasoning strategies (general-purpose reasoning in ZSL, reasoning traces in SFT, and Reinforcement Learning with Verifiable Reward (RLVR) with novel reward functions) affect Text2SQL performance across four benchmarks. We show that partial scoring rewards, computed via SQL execution, are crucial for guiding models even when outputs are not fully correct. These fine-grained signals lead to consistently better Text2SQL outcomes. Small LLMs benefit most from reasoning-aware SFT and RL, with the 14B Qwen-Coder-2.5 surpassing 400B+ models on challenging datasets like BIRD.

cs.LG↗

arXiv2025

Optical control of the spin-Hall effect in a two-dimensional hole gas

Relativistic effects influence the motion of charged particles in solids by intertwining spin and momentum. The resulting phenomena exhibit rich and intriguing properties that can unveil radically new quantum devices. In this context, the two-dimensional hole gas formed in group IV heterostructures is a particularly promising platform, owning to a notable spin-orbit coupling. However, the exploitation of spin-momentum locking and precise manipulation of spin currents has remained elusive thus far. Here we use the modulation-doping technique to break inversion symmetry at novel Ge1-xSnx/Ge interfaces and explore spin-orbit phenomena in the emergent Rashba-coupled hole gases. Magneto-optical investigations demonstrate the unusual establishment of a staggered band alignment with carrier lifetime in the ns range. Optical spin orientation is then leveraged to directly inject spin-polarized currents in the Rashba-split 2D gas. Spin-to-charge conversion is shown to genuinely occur at the staggered gap through the inverse spin-Hall effect. This provides unprecedented access to low-order contributions of the spin-orbit Hamiltonian. Moreover, it leads to the startling demonstration that the spin Hall angle can be optically controlled by modifying the Rashba coupling through the photoexcitation density. Ge1-xSnx quantum wells thus offer innovative solutions and functionalities stemming from their unique spin-dependent properties and intriguing quantum phenomena at the crossroad between transport and photonic realms.

cond-mat.mes-hall↗

arXiv2024

Optimizing Diffusion Models for Joint Trajectory Prediction and Controllable Generation

Diffusion models are promising for joint trajectory prediction and controllable generation in autonomous driving, but they face challenges of inefficient inference steps and high computational demands. To tackle these challenges, we introduce Optimal Gaussian Diffusion (OGD) and Estimated Clean Manifold (ECM) Guidance. OGD optimizes the prior distribution for a small diffusion time $T$ and starts the reverse diffusion process from it. ECM directly injects guidance gradients to the estimated clean manifold, eliminating extensive gradient backpropagation throughout the network. Our methodology streamlines the generative process, enabling practical applications with reduced computational overhead. Experimental validation on the large-scale Argoverse 2 dataset demonstrates our approach's superior performance, offering a viable solution for computationally efficient, high-quality joint trajectory prediction and controllable generation for autonomous driving. Our project webpage is at https://yixiaowang7.github.io/OptTrajDiff_Page/.

cs.CV↗

arXiv2023

Class Balanced Dynamic Acquisition for Domain Adaptive Semantic Segmentation using Active Learning

Domain adaptive active learning is leading the charge in label-efficient training of neural networks. For semantic segmentation, state-of-the-art models jointly use two criteria of uncertainty and diversity to select training labels, combined with a pixel-wise acquisition strategy. However, we show that such methods currently suffer from a class imbalance issue which degrades their performance for larger active learning budgets. We then introduce Class Balanced Dynamic Acquisition (CBDA), a novel active learning method that mitigates this issue, especially in high-budget regimes. The more balanced labels increase minority class performance, which in turn allows the model to outperform the previous baseline by 0.6, 1.7, and 2.4 mIoU for budgets of 5%, 10%, and 20%, respectively. Additionally, the focus on minority classes leads to improvements of the minimum class performance of 0.5, 2.9, and 4.6 IoU respectively. The top-performing model even exceeds the fully supervised baseline, showing that a more balanced label than the entire ground truth can be beneficial.

cs.CV↗

arXiv2023

On permutation symmetries in Bayesian neural network posteriors: a variational perspective

The elusive nature of gradient-based optimization in neural networks is tied to their loss landscape geometry, which is poorly understood. However recent work has brought solid evidence that there is essentially no loss barrier between the local solutions of gradient descent, once accounting for weight-permutations that leave the network's computation unchanged. This raises questions for approximate inference in Bayesian neural networks (BNNs), where we are interested in marginalizing over multiple points in the loss landscape. In this work, we first extend the formalism of marginalized loss barrier and solution interpolation to BNNs, before proposing a matching algorithm to search for linearly connected solutions. This is achieved by aligning the distributions of two independent approximate Bayesian solutions with respect to permutation matrices. We build on the results of Ainsworth et al. (2023), reframing the problem as a combinatorial optimization one, using an approximation to the sum of bilinear assignment problem. We then experiment on a variety of architectures and datasets, finding nearly zero marginalized loss barriers for linearly connected solutions.

stat.ML↗

arXiv2023

Simulating Cardiac Fluid Dynamics in the Human Heart

Cardiac fluid dynamics fundamentally involves interactions between complex blood flows and the structural deformations of the muscular heart walls and the thin, flexible valve leaflets. There has been longstanding scientific, engineering, and medical interest in creating mathematical models of the heart that capture, explain, and predict these fluid-structure interactions. However, existing computational models that account for interactions among the blood, the actively contracting myocardium, and the cardiac valves are limited in their abilities to predict valve performance, resolve fine-scale flow features, or use realistic descriptions of tissue biomechanics. Here we introduce and benchmark a comprehensive mathematical model of cardiac fluid dynamics in the human heart. A unique feature of our model is that it incorporates biomechanically detailed descriptions of all major cardiac structures that are calibrated using tensile tests of human tissue specimens to reflect the heart's microstructure. Further, it is the first fluid-structure interaction model of the heart that provides anatomically and physiologically detailed representations of all four cardiac valves. We demonstrate that this integrative model generates physiologic dynamics, including realistic pressure-volume loops that automatically capture isovolumetric contraction and relaxation, and predicts fine-scale flow features. None of these outputs are prescribed; instead, they emerge from interactions within our comprehensive description of cardiac physiology. Such models can serve as tools for predicting the impacts of medical devices or clinical interventions. They also can serve as platforms for mechanistic studies of cardiac pathophysiology and dysfunction, including congenital defects, cardiomyopathies, and heart failure, that are difficult or impossible to perform in patients.

q-bio.TO↗

arXiv2023

Continuous-Time Functional Diffusion Processes

We introduce Functional Diffusion Processes (FDPs), which generalize score-based diffusion models to infinite-dimensional function spaces. FDPs require a new mathematical framework to describe the forward and backward dynamics, and several extensions to derive practical training objectives. These include infinite-dimensional versions of Girsanov theorem, in order to be able to compute an ELBO, and of the sampling theorem, in order to guarantee that functional evaluations in a countable set of points are equivalent to infinite-dimensional functions. We use FDPs to build a new breed of generative models in function spaces, which do not require specialized network architectures, and that can work with any kind of continuous data. Our results on real data show that FDPs achieve high-quality image generation, using a simple MLP architecture with orders of magnitude fewer parameters than existing diffusion models.

cs.LG↗

arXiv2022

How Much is Enough? A Study on Diffusion Times in Score-based Generative Models

Score-based diffusion models are a class of generative models whose dynamics is described by stochastic differential equations that map noise into data. While recent works have started to lay down a theoretical foundation for these models, an analytical understanding of the role of the diffusion time T is still lacking. Current best practice advocates for a large T to ensure that the forward dynamics brings the diffusion sufficiently close to a known and simple noise distribution; however, a smaller value of T should be preferred for a better approximation of the score-matching objective and higher computational efficiency. Starting from a variational interpretation of diffusion models, in this work we quantify this trade-off, and suggest a new method to improve quality and efficiency of both training and sampling, by adopting smaller diffusion times. Indeed, we show how an auxiliary model can be used to bridge the gap between the ideal and the simulated forward dynamics, followed by a standard reverse diffusion process. Empirical results support our analysis; for image data, our method is competitive w.r.t. the state-of-the-art, according to standard sample quality metrics and log-likelihood.

stat.ML↗

arXiv2022

A sharp interface Lagrangian-Eulerian method for flexible-body fluid-structure interaction

This paper introduces a sharp-interface approach to simulating fluid-structure interaction involving flexible bodies described by general nonlinear material models and across a broad range of mass density ratios. This new flexible-body immersed Lagrangian-Eulerian (ILE) approach incorporates the geometrical and domain solution flexibility of the immersed boundary (IB) method with an accuracy comparable to body-fitted approaches that sharply resolve flows and stresses up to the fluid-structure interface. Unlike many IB methods, our ILE formulation uses distinct momentum equations for the fluid and solid subregions with a Dirichlet-Neumann coupling strategy that connects fluid and solid subproblems through simple interface conditions. We use a penalty method involving two representations of the fluid-structure interface. These two representations are connected by approximate Lagrange multiplier forces that impose kinematic interface conditions. This approach also enables the use of multi-rate time stepping, which allows us to take different time step sizes for the fluid and structure subproblems. Our fluid solver relies on an immersed interface method for discrete surfaces to impose stress jump conditions along complex interfaces. The dynamics of the volumetric structural mesh are determined using a standard finite element approach to large-deformation nonlinear elasticity via a nearly incompressible solid mechanics formulation. This formulation also readily accommodates compressible structures for cases in which part of the solid boundary does not contact the incompressible fluid. Comparisons are made with computational and experimental benchmarks. We also demonstrate the capabilities of this methodology by applying it to model the transport and capture of a cylindrical blood clot in an inferior vena cava filter.

math.NA↗

arXiv2021

Model Selection for Bayesian Autoencoders

We develop a novel method for carrying out model selection for Bayesian autoencoders (BAEs) by means of prior hyper-parameter optimization. Inspired by the common practice of type-II maximum likelihood optimization and its equivalence to Kullback-Leibler divergence minimization, we propose to optimize the distributional sliced-Wasserstein distance (DSWD) between the output of the autoencoder and the empirical data distribution. The advantages of this formulation are that we can estimate the DSWD based on samples and handle high-dimensional problems. We carry out posterior estimation of the BAE parameters via stochastic gradient Hamiltonian Monte Carlo and turn our BAE into a generative model by fitting a flexible Dirichlet mixture model in the latent space. Consequently, we obtain a powerful alternative to variational autoencoders, which are the preferred choice in modern applications of autoencoders for representation learning with uncertainty. We evaluate our approach qualitatively and quantitatively using a vast experimental campaign on a number of unsupervised learning tasks and show that, in small-data regimes where priors matter, our approach provides state-of-the-art results, outperforming multiple competitive baselines.

stat.ML↗

arXiv2021

Enhancing human bodies with extra robotic arms and fingers: The Neural Resource Allocation Problem

The emergence of robot-based body augmentation promises exciting innovations that will inform robotics, human-machine interaction, and wearable electronics. Even though augmentative devices like extra robotic arms and fingers in many ways build on restorative technologies, they introduce unique challenges for bidirectional human-machine collaboration. Can humans adapt and learn to operate a new limb collaboratively with their biological limbs without sacrificing their physical abilities? To successfully achieve robotic body augmentation, we need to ensure that by giving a person an additional (artificial) limb, we are not in fact trading off an existing (biological) one. In this manuscript, we introduce the "Neural Resource Allocation" problem, which distinguishes body augmentation from existing robotics paradigms such as teleoperation and prosthetics. We discuss how to allow the effective and effortless voluntary control of augmentative devices without compromising the voluntary control of the biological body. In reviewing the relevant literature on extra robotic fingers and limbs we critically assess the range of potential solutions available for the "Neural Resource Allocation" problem. For this purpose, we combine multiple perspectives from engineering and neuroscience with considerations from human-machine interaction, sensory-motor integration, ethics and law. Altogether we aim to define common foundations and operating principles for the successful implementation of motor augmentation.

cs.RO↗

arXiv2020

All You Need is a Good Functional Prior for Bayesian Deep Learning

The Bayesian treatment of neural networks dictates that a prior distribution is specified over their weight and bias parameters. This poses a challenge because modern neural networks are characterized by a large number of parameters, and the choice of these priors has an uncontrolled effect on the induced functional prior, which is the distribution of the functions obtained by sampling the parameters from their prior distribution. We argue that this is a hugely limiting aspect of Bayesian deep learning, and this work tackles this limitation in a practical and effective way. Our proposal is to reason in terms of functional priors, which are easier to elicit, and to "tune" the priors of neural network parameters in a way that they reflect such functional priors. Gaussian processes offer a rigorous framework to define prior distributions over functions, and we propose a novel and robust framework to match their prior with the functional prior of neural networks based on the minimization of their Wasserstein distance. We provide vast experimental evidence that coupling these priors with scalable Markov chain Monte Carlo sampling offers systematically large performance improvements over alternative choices of priors and state-of-the-art approximate Bayesian deep learning approaches. We consider this work a considerable step in the direction of making the long-standing challenge of carrying out a fully Bayesian treatment of neural networks, including convolutional neural networks, a concrete possibility.

stat.ML↗

arXiv2020

Magneto-optical determination of the carrier lifetime in coherent Ge(1-x)Sn(x)/Ge heterostructures

We present a magneto-optical study of the carrier dynamics in compressively strained Ge(1-x)Sn(x) films having Sn compositions up to 10% epitaxially grown on blanket Ge on Si (001) virtual substrates. We leverage the Hanle effect under steady-state excitation to study the spin-dependent optical transitions in presence of an external magnetic field. This allowed us to obtain direct access to the dynamics of the optically-induced carrier population. Our approach singled out that at cryogenic temperatures the effective lifetime of the photogenerated carriers in coherent Ge(1-x)Sn(x) occurs in the sub-ns time scale. Supported by a model estimate of the radiative lifetime, our measurements indicate that carrier recombination is dominated by non-radiative processes. Our results thus provide central information to advance the fundamental understanding of carrier kinetics in this novel direct-gap group-IV material system. Such knowledge can be a stepping stone in the quest for the implementation of Ge(1-x)Sn(x)-based functional devices.

physics.optics↗