arXiv subjects

Francesco Rinaldi

Publications and source records attributed to Francesco Rinaldi.

At least 19 recordsLinked to original sources

arXiv2026

Extrapolation-based Direct Search for Nonsmooth Stochastic Zeroth-Order Optimization

We propose and analyze a stochastic direct-search method for unconstrained zeroth-order minimization of locally Lipschitz, possibly nonsmooth, objectives. The method combines random polling directions with a stochastic extrapolating line search based on a sufficient-decrease test of order $p$. Under conditional accuracy assumptions on the stochastic estimates, we prove almost-sure convergence to Clarke stationary points. We further establish an expected iteration complexity bound. Specifically, using a supermartingale stopping-time argument, we prove that $\mathcal O\left( \max\left\{ r^{-p}, \varepsilon^{-p/(p-1)} \right\} \right) $ iterations are sufficient in expectation to reach an $(r,\varepsilon)$-Goldstein stationary point. Moreover, we derive a corresponding expected tested-point complexity bound of order $\mathcal O\bigl(\varepsilon^{1-n} \max\{r^{-p},\varepsilon^{-p/(p-1)}\}\bigr)$. To the best of our knowledge, this is the first convergence and expected-complexity analysis for an extrapolation-based direct-search method in a nonsmooth stochastic setting. Numerical experiments on a DFO benchmark suite highlight competitive performance against well-established stochastic direct-search methods.

math.OC↗

arXiv2026

Constrained Variable Projection for Structured Problems

Variable projection is a classical technique for separable nonlinear least-squares problems, in which variables that enter linearly are eliminated exactly, yielding a reduced nonlinear problem. By expressing this framework as a particular instance of a broader class of bilevel optimization problems, we develop a constrained variable-projection framework for data-science models, where the remaining variables are subject to convex constraints and the eliminated variables arise from a lower-level least-squares problem. In particular, by interpreting variable projection as a collapsed bilevel optimization problem, we derive exact reduced-gradient formulas compatible with automatic differentiation and propose a conditional-gradient algorithm for the resulting constrained reduced problem. We establish convergence guarantees under standard smoothness and compactness assumptions, and discuss extensions to structured lower-level variables. Numerical experiments on sparse autoencoding, dictionary learning, blind deconvolution, and few-shot learning suggest that the method can improve wall-clock efficiency and data efficiency relative to natural joint-optimization baselines.

math.OC↗

arXiv2026

Iteration Complexity of Frank-Wolfe and Its Variants for Bilevel Optimization

We study Frank-Wolfe (FW) methods for constrained bilevel optimization when the lower-level problem is solved only approximately, yielding biased and inexact hypergradients. We analyze inexact variants of vanilla FW as well as away-step and pairwise FW, and provide convergence rates in the nonconvex setting under gradient errors. By combining these results with recent bounds on hypergradient errors from iterative and approximate implicit differentiation, we derive overall iteration complexity guarantees for bilevel FW. Experiments on two real-world applications validate the theory and demonstrate practical effectiveness.

math.OC↗

arXiv2026

The Maximum Clique Problem under Adversarial Uncertainty: a min-max approach

We analyze the problem of identifying large cliques in graphs that are affected by adversarial uncertainty. More specifically, we consider a new formulation, namely the adversarial maximum clique problem, which extends the classical maximum-clique problem to graphs with edges strategically perturbed by an adversary. The proposed mathematical model is thus formulated as a two-player zero-sum game between a clique seeker and an opposing agent. Inspired by regularized continuous reformulations of the maximum-clique problem, we derive a penalized continuous formulation leading to a nonconvex and nonsmooth optimization problem. We further introduce the notion of stable global solutions, namely points remaining optimal under small perturbations of the penalty parameters, and prove an equivalence between stable global solutions of the continuous reformulation and largest cliques that are common to all the adversarially perturbed graphs. In order to solve the given nonsmooth problem, we develop a first-order and projection-free algorithm based on generalized subdifferential calculus in the sense of Clarke and Goldstein, and establish global sublinear convergence rates for it. Finally, we report numerical experiments on benchmark instances showing that the proposed method efficiently detects large common cliques.

math.OC↗

arXiv2025

The Effects of Remote Working on Scientific Collaboration and Impact

The COVID-19 pandemic shifted academic collaboration from in-person to remote interactions. This study explores, for the first time, the effects on scientific collaborations and impact of such a shift, comparing research output before, during, and after the pandemic. Using large-scale bibliometric data, we track the evolution of collaboration networks and the resulting impact of research over time. Our findings are twofold: first, the geographic distribution of collaborations significantly shifted, with a notable increase in cross-border partnerships after 2020, indicating a reduction in the constraints of geographic proximity. Second, despite the expansion of collaboration networks, there was a concerning decline in citation impact, suggesting that the absence of spontaneous in-person interactions-which traditionally foster deep discussions and idea exchange-negatively affected research quality. As hybrid work models in academia gain traction, this study highlights the need for universities and research organizations to carefully consider the balance between remote and in-person engagement.

physics.soc-ph↗

arXiv2025

Sequential test sampling for stochastic derivative-free optimization

In many derivative-free optimization algorithms, a sufficient decrease condition decides whether to accept a trial step in each iteration. This condition typically requires that the potential objective function value decrease of the trial step, i.e., the true reduction in the objective function value that would be achieved by moving from the current point to the trial point, be larger than a multiple of the squared stepsize. When the objective function is stochastic, evaluating such a condition accurately can require a large estimation cost. In this paper, we frame the evaluation of the sufficient decrease condition in a stochastic setting as a hypothesis test problem and solve it through a sequential hypothesis test. The two hypotheses considered in the problem correspond to accepting or rejecting the trial step. This test sequentially collects noisy sample observations of the potential decrease until their sum crosses either a lower or an upper boundary depending on the noise variance and the stepsize. When the noise of observations is Gaussian, we derive a novel sample size result, showing that the effort to evaluate the condition explicitly depends on the potential decrease, and that the sequential test terminates early whenever the sufficient decrease condition is away from satisfaction. Furthermore, when the potential decrease is~$\Theta(\delta^r)$ for some~$r\in(0,2]$, the expected sample size decreases from~$\Theta(\delta^{-4})$ to~$O(\delta^{-2-r})$. We apply this sequential test sampling framework to probabilistic-descent direct search. To analyze its convergence rate, we extend a renewal-reward supermartingale-based convergence rate analysis framework to an arbitrary probability threshold. By doing so, we are able to show that probabilistic-descent direct search has an iteration complexity of $O(n/\epsilon^2)$ for gradient norm...

math.OC↗

arXiv2025

An Efficient Network-aware Direct Search Method for Influence Maximization

Influence Maximization (IM) is a pivotal concept in social network analysis, involving the identification of influential nodes within a network to maximize the number of influenced nodes, and has a wide variety of applications that range from viral marketing and information dissemination to public health campaigns. IM can be modeled as a combinatorial optimization problem with a black-box objective function, where the goal is to select $B$ seed nodes that maximize the expected influence spread. Direct search methods, which do not require gradient information, are well-suited for such problems. Unlike gradient-based approaches, direct search algorithms, in fact, only evaluate the objective function at a suitably chosen set of trial points around the current solution to guide the search process. However, these methods often suffer from scalability issues due to the high cost of function evaluations, especially when applied to combinatorial problems like IM. This work, therefore, proposes the Network-aware Direct Search (NaDS) method, an innovative direct search approach that integrates the network structure into its neighborhood formulation and is used to tackle a mixed-integer programming formulation of the IM problem, the so-called General Information Propagation model. We tested our method on large-scale networks, comparing it to existing state-of-the-art approaches for the IM problem, including direct search methods and various greedy techniques and heuristics. The results of the experiments empirically confirm the assumptions underlying NaDS, demonstrating that exploiting the graph structure of the IM problem in the algorithmic framework can significantly improve its computational efficiency in the considered context.

cs.SI↗

arXiv2025

Fast and Simple Multiclass Data Segmentation: An Eigendecomposition and Projection-Free Approach

Graph-based machine learning has seen an increased interest over the last decade with many connections to other fields of applied mathematics. Learning based on partial differential equations, such as the phase-field Allen-Cahn equation, allows efficient handling of semi-supervised learning approaches on graphs. The numerical solution of the graph Allen-Cahn equation via a convexity splitting or the Merriman-Bence-Osher (MBO) scheme, albeit being a widely used approach, requires the calculation of a graph Laplacian eigendecomposition and repeated projections over the unit simplex to maintain valid partitions. The computational efficiency of those methods is hence limited by those two bottlenecks in practice, especially when dealing with large-scale instances. In order to overcome these limitations, we propose a new framework combining a novel penalty-based reformulation of the segmentation problem, which ensures valid partitions (i.e., binary solutions) for appropriate parameter choices, with an eigendecomposition and projection-free optimization scheme, which has a small per-iteration complexity (by relying primarily on sparse matrix-vector products) and guarantees good convergence properties. Experiments on synthetic and real-world datasets related to data segmentation in networks and images demonstrate that the proposed framework achieves comparable or better accuracy than the CS and MBO methods while being significantly faster, particularly for large-scale problems.

math.NA↗

arXiv2025

Exploring the Potential of Bilevel Optimization for Calibrating Neural Networks

Handling uncertainty is critical for ensuring reliable decision-making in intelligent systems. Modern neural networks are known to be poorly calibrated, resulting in predicted confidence scores that are difficult to use. This article explores improving confidence estimation and calibration through the application of bilevel optimization, a framework designed to solve hierarchical problems with interdependent optimization levels. A self-calibrating bilevel neural-network training approach is introduced to improve a model's predicted confidence scores. The effectiveness of the proposed framework is analyzed using toy datasets, such as Blobs and Spirals, as well as more practical simulated datasets, such as Blood Alcohol Concentration (BAC). It is compared with a well-known and widely used calibration strategy, isotonic regression. The reported experimental results reveal that the proposed bilevel optimization approach reduces the calibration error while preserving accuracy.

cs.LG↗

arXiv2024

What we should learn from pandemic publishing

Authors of COVID-19 papers produced during the pandemic were overwhelmingly not subject matter experts. Such a massive inflow of scholars from different expertise areas is both an asset and a potential problem. Domain-informed scientific collaboration is the key to preparing for future crises.

physics.soc-ph↗

arXiv2024

Probabilistic Iterative Hard Thresholding for Sparse Learning

For statistical modeling wherein the data regime is unfavorable in terms of dimensionality relative to the sample size, finding hidden sparsity in the ground truth can be critical in formulating an accurate statistical model. The so-called "l0 norm" which counts the number of non-zero components in a vector, is a strong reliable mechanism of enforcing sparsity when incorporated into an optimization problem for minimizing the fit of a given model to a set of observations. However, in big data settings wherein noisy estimates of the gradient must be evaluated out of computational necessity, the literature is scant on methods that reliably converge. In this paper we present an approach towards solving expectation objective optimization problems with cardinality constraints. We prove convergence of the underlying stochastic process, and demonstrate the performance on two Machine Learning problems.

math.OC↗

arXiv2023

Projection free methods on product domains

Projection-free block-coordinate methods avoid high computational cost per iteration and at the same time exploit the particular problem structure of product domains. Frank-Wolfe-like approaches rank among the most popular ones of this type. However, as observed in the literature, there was a gap between the classical Frank-Wolfe theory and the block-coordinate case. Moreover, most of previous research concentrated on convex objectives. This study now deals also with the non-convex case and reduces above-mentioned theory gap, in combining a new, fully developed convergence theory with novel active set identification results which ensure that inherent sparsity of solutions can be exploited in an efficient way. Preliminary numerical experiments seem to justify our approach and also show promising results for obtaining global solutions in the non-convex case.

math.OC↗

arXiv2023

Laplacian-based Semi-Supervised Learning in Multilayer Hypergraphs by Coordinate Descent

Graph Semi-Supervised learning is an important data analysis tool, where given a graph and a set of labeled nodes, the aim is to infer the labels to the remaining unlabeled nodes. In this paper, we start by considering an optimization-based formulation of the problem for an undirected graph, and then we extend this formulation to multilayer hypergraphs. We solve the problem using different coordinate descent approaches and compare the results with the ones obtained by the classic gradient descent method. Experiments on synthetic and real-world datasets show the potential of using coordinate descent methods with suitable selection rules.

cs.LG↗

arXiv2023

Inexact Direct-Search Methods for Bilevel Optimization Problems

In this work, we introduce new direct search schemes for the solution of bilevel optimization (BO) problems. Our methods rely on a fixed accuracy black box oracle for the lower-level problem, and deal both with smooth and potentially nonsmooth true objectives. We thus analyze for the first time in the literature direct search schemes in these settings, giving convergence guarantees to approximate stationary points, as well as complexity bounds in the smooth case. We also propose the first adaptation of mesh adaptive direct search schemes for BO. Some preliminary numerical results on a standard set of bilevel optimization problems show the effectiveness of our new approaches.

math.OC↗

arXiv2023

Relax and penalize: a new bilevel approach to mixed-binary hyperparameter optimization

In recent years, bilevel approaches have become very popular to efficiently estimate high-dimensional hyperparameters of machine learning models. However, to date, binary parameters are handled by continuous relaxation and rounding strategies, which could lead to inconsistent solutions. In this context, we tackle the challenging optimization of mixed-binary hyperparameters by resorting to an equivalent continuous bilevel reformulation based on an appropriate penalty term. We propose an algorithmic framework that, under suitable assumptions, is guaranteed to provide mixed-binary solutions. Moreover, the generality of the method allows to safely use existing continuous bilevel solvers within the proposed framework. We evaluate the performance of our approach for two specific machine learning problems, i.e., the estimation of the group-sparsity structure in regression problems and the data distillation problem. The reported results show that our method is competitive with state-of-the-art approaches based on relaxation and rounding

cs.LG↗

arXiv2023

Stochastic trust-region and direct-search methods: A weak tail bound condition and reduced sample sizing

Using tail bounds, we introduce a new probabilistic condition for function estimation in stochastic derivative-free optimization which leads to a reduction in the number of samples and eases algorithmic analyses. Moreover, we develop simple stochastic direct-search and trust-region methods for the optimization of a potentially non-smooth function whose values can only be estimated via stochastic observations. For trial points to be accepted, these algorithms require the estimated function values to yield a sufficient decrease measured in terms of a power larger than 1 of the algoritmic stepsize. Our new tail bound condition is precisely imposed on the reduction estimate used to achieve such a sufficient decrease. This condition allows us to select the stepsize power used for sufficient decrease in such a way to reduce the number of samples needed per iteration. In previous works, the number of samples necessary for global convergence at every iteration $k$ of this type of algorithms was $O(Δ_{k}^{-4})$, where $Δ_k$ is the stepsize or trust-region radius. However, using the new tail bound condition, and under mild assumptions on the noise, one can prove that such a number of samples is only $O(Δ_k^{-2 - \varepsilon})$, where $\varepsilon > 0$ can be made arbitrarily small by selecting the power of the stepsize in the sufficient decrease test arbitrarily close to $1$. The global convergence properties of the stochastic direct-search and trust-region algorithms are established under the new tail bound condition.

math.OC↗

arXiv2023

Learning the Right Layers: a Data-Driven Layer-Aggregation Strategy for Semi-Supervised Learning on Multilayer Graphs

Clustering (or community detection) on multilayer graphs poses several additional complications with respect to standard graphs as different layers may be characterized by different structures and types of information. One of the major challenges is to establish the extent to which each layer contributes to the cluster assignment in order to effectively take advantage of the multilayer structure and improve upon the classification obtained using the individual layers or their union. However, making an informed a-priori assessment about the clustering information content of the layers can be very complicated. In this work, we assume a semi-supervised learning setting, where the class of a small percentage of nodes is initially provided, and we propose a parameter-free Laplacian-regularized model that learns an optimal nonlinear combination of the different layers from the available input labels. The learning algorithm is based on a Frank-Wolfe optimization scheme with inexact gradient, combined with a modified Label Propagation iteration. We provide a detailed convergence analysis of the algorithm and extensive experiments on synthetic and real-world datasets, showing that the proposed method compares favourably with a variety of baselines and outperforms each individual layer when used in isolation.

cs.LG↗

arXiv2023

Collaboration and topic switches in science

Collaboration is a key driver of science and innovation. Mainly motivated by the need to leverage different capacities and expertise to solve a scientific problem, collaboration is also an excellent source of information about the future behavior of scholars. In particular, it allows us to infer the likelihood that scientists choose future research directions via the intertwined mechanisms of selection and social influence. Here we thoroughly investigate the interplay between collaboration and topic switches. We find that the probability for a scholar to start working on a new topic increases with the number of previous collaborators, with a pattern showing that the effects of individual collaborators are not independent. The higher the productivity and the impact of authors, the more likely their coworkers will start working on new topics. The average number of coauthors per paper is also inversely related to the topic switch probability, suggesting a dilution of this effect as the number of collaborators increases.

physics.soc-ph↗