arXiv subjects

Ioannis Kordonis

Publications and source records attributed to Ioannis Kordonis.

11 recordsLinked to original sources

arXiv2026

Optimal Transport for Handwritten Text Recognition in a Low-Resource Regime

Handwritten Text Recognition (HTR) is a task of central importance in the field of document image understanding. State-of-the-art methods for HTR require the use of extensive annotated sets for training, making them impractical for low-resource domains like historical archives or limited-size modern collections. This paper introduces a novel framework that, unlike the standard HTR model paradigm, can leverage mild prior knowledge of lexical characteristics; this is ideal for scenarios where labeled data are scarce. We propose an iterative bootstrapping approach that aligns visual features extracted from unlabeled images with semantic word representations using Optimal Transport (OT). Starting with a minimal set of labeled examples, the framework iteratively matches word images to text labels, generates pseudo-labels for high-confidence alignments, and retrains the recognizer on the growing dataset. Numerical experiments demonstrate that our iterative visual-semantic alignment scheme significantly improves recognition accuracy on low-resource HTR benchmarks.

cs.CV↗

arXiv2026

Diffusion-Based Scenario Tree Generation for Multivariate Time Series Prediction and Multistage Stochastic Optimization

Stochastic forecasting is critical for efficient decision-making in uncertain systems, such as energy markets and finance, where estimating the full distribution of future scenarios is essential. We propose Diffusion Scenario Tree (DST), a general framework for constructing scenario trees using diffusion-based probabilistic forecasting models to provide a structured model of system evolution for control tasks. DST recursively samples future trajectories and organizes them into a tree via clustering, ensuring non-anticipativity (decisions depending only on observed history) at each stage, offering a superior representation of uncertainty compared to using predictive models solely for forecasting system evolution. We integrate DST into Model Predictive Control (MPC) and evaluate it on energy arbitrage in New York State's day-ahead electricity market. Experimental results show that our approach significantly outperforms the same optimization algorithms that use scenario trees generated by more conventional models. Furthermore, using DST for stochastic optimization yields more efficient decision policies by better handling uncertainty than deterministic and stochastic MPC variants using the same diffusion-based forecaster, and simple Model-Free Reinforcement Learning (RL) baselines.

cs.LG↗

arXiv2025

Shared-Weights Extender and Gradient Voting for Neural Network Expansion

Expanding neural networks during training is a promising way to augment capacity without retraining larger models from scratch. However, newly added neurons often fail to adjust to a trained network and become inactive, providing no contribution to capacity growth. We propose the Shared-Weights Extender (SWE), a novel method explicitly designed to prevent inactivity of new neurons by coupling them with existing ones for smooth integration. In parallel, we introduce the Steepest Voting Distributor (SVoD), a gradient-based method for allocating neurons across layers during deep network expansion. Our extensive benchmarking on four datasets shows that our method can effectively suppress neuron inactivity and achieve better performance compared to other expanding methods and baselines.

cs.LG↗

arXiv2025

Diffusion-Based Forecasting for Uncertainty-Aware Model Predictive Control

We propose Diffusion-Informed Model Predictive Control (D-I MPC), a generic framework for uncertainty-aware prediction and decision-making in partially observable stochastic systems by integrating diffusion-based time series forecasting models in Model Predictive Control algorithms. In our approach, a diffusion-based time series forecasting model is used to probabilistically estimate the evolution of the system's stochastic components. These forecasts are then incorporated into MPC algorithms to estimate future trajectories and optimize action selection under the uncertainty of the future. We evaluate the framework on the task of energy arbitrage, where a Battery Energy Storage System participates in the day-ahead electricity market of the New York state. Experimental results indicate that our model-based approach with a diffusion-based forecaster significantly outperforms both implementations with classical forecasting methods and model-free reinforcement learning baselines.

cs.LG↗

arXiv2023

Matrix Factorization in Tropical and Mixed Tropical-Linear Algebras

Matrix Factorization (MF) has found numerous applications in Machine Learning and Data Mining, including collaborative filtering recommendation systems, dimensionality reduction, data visualization, and community detection. Motivated by the recent successes of tropical algebra and geometry in machine learning, we investigate two problems involving matrix factorization over the tropical algebra. For the first problem, Tropical Matrix Factorization (TMF), which has been studied already in the literature, we propose an improved algorithm that avoids many of the local optima. The second formulation considers the approximate decomposition of a given matrix into the product of three matrices where a usual matrix product is followed by a tropical product. This formulation has a very interesting interpretation in terms of the learning of the utility functions of multiple users. We also present numerical results illustrating the effectiveness of the proposed algorithms, as well as an application to recommendation systems with promising results.

cs.LG↗

arXiv2023

Revisiting Tropical Polynomial Division: Theory, Algorithms and Application to Neural Networks

Tropical geometry has recently found several applications in the analysis of neural networks with piecewise linear activation functions. This paper presents a new look at the problem of tropical polynomial division and its application to the simplification of neural networks. We analyze tropical polynomials with real coefficients, extending earlier ideas and methods developed for polynomials with integer coefficients. We first prove the existence of a unique quotient-remainder pair and characterize the quotient in terms of the convex bi-conjugate of a related function. Interestingly, the quotient of tropical polynomials with integer coefficients does not necessarily have integer coefficients. Furthermore, we develop a relationship of tropical polynomial division with the computation of the convex hull of unions of convex polyhedra and use it to derive an exact algorithm for tropical polynomial division. An approximate algorithm is also presented, based on an alternation between data partition and linear programming. We also develop special techniques to divide composite polynomials, described as sums or maxima of simpler ones. Finally, we present some numerical results to illustrate the efficiency of the algorithms proposed, using the MNIST handwritten digit and CIFAR-10 datasets.

cs.LG↗

arXiv2021

Dynamic Games of Social Distancing during an Epidemic: Analysis of Asymmetric Solutions

Individual behaviors play an essential role in the dynamics of transmission of infectious diseases, including COVID--19. This paper studies a dynamic game model that describes the social distancing behaviors during an epidemic, assuming a continuum of players and individual infection dynamics. The evolution of the players' infection states follows a variant of the well-known SIR dynamics. We assume that the players are not sure about their infection state, and thus they choose their actions based on their individually perceived probabilities of being susceptible, infected or removed. The cost of each player depends both on her infection state and on the contact with others. We prove the existence of a Nash equilibrium and characterize Nash equilibria using nonlinear complementarity problems. We then exploit some monotonicity properties of the optimal policies to obtain a reduced-order characterization for Nash equilibrium and reduce its computation to the solution of a low-dimensional optimization problem. It turns out that, even in the symmetric case, where all the players have the same parameters, players may have very different behaviors. We finally present some numerical studies that illustrate this interesting phenomenon and investigate the effects of several parameters, including the players' vulnerability, the time horizon, and the maximum allowed actions, on the optimal policies and the players' costs.

cs.GT↗

arXiv2021

Nash Social Distancing Games with Equity Constraints: How Inequality Aversion Affects the Spread of Epidemics

In this paper, we present a game-theoretic model describing voluntary social distancing during the spread of an epidemic. The payoffs of the agents depend on the social distancing they practice and on the probability of getting infected. We consider two types of agents, the non-vulnerable agents who have a small cost if they get infected, and the vulnerable agents who have a higher cost. For the modeling of the epidemic outbreak, we consider a variant of the SIR (Susceptible-Infected-Removed) model involving populations of susceptible, infected, and removed persons of vulnerable and non-vulnerable types. The Nash equilibria of this social distancing game are studied. The main contribution of this work is the analysis of the case where the players, desiring to achieve a low social inequality, pose a bound on the variance of the payoffs. In this case, we introduce and characterize a notion of Generalized Nash Equilibrium (GNE) for games with a continuum of players. Through numerical studies, we show that inequality constraints result in a slower spread of the epidemic and an improved cost for the vulnerable players. Furthermore, it is possible that inequality constraints are beneficial for non-vulnerable players as well.

cs.GT↗

arXiv2021

Personalized Pareto-Improving Pricing-and-Routing Schemes for Near-Optimum Freight Routing: An Alternative Approach to Congestion Pricing

We design a coordination mechanism for truck drivers that uses pricing-and-routing schemes that can help alleviate traffic congestion in a general transportation network. We consider the user heterogeneity in Value-Of-Time (VOT) by adopting a multi-class model with stochastic Origin-Destination (OD) demands for the truck drivers. The main characteristic of the mechanism is that the coordinator asks the truck drivers to declare their desired OD pair and pick their individual VOT from a set of $N$ available options, and guarantees that the resulting pricing-and-routing scheme is Pareto-improving, i.e. every truck driver will be better-off compared to the User Equilibrium (UE) and that every truck driver will have an incentive to truthfully declare his/her VOT, while leading to a revenue-neutral (budget balanced) on average mechanism. This approach enables us to design personalized (VOT-based) pricing-and-routing schemes. We show that the Optimum Pricing Scheme (OPS) can be calculated by solving a nonconvex optimization problem. To improve computational efficiency, we propose an Approximately Optimum Pricing Scheme (AOPS) and prove that it satisfies the aforementioned properties. Both pricing-and-routing schemes are compared to the Congestion Pricing with Uniform Revenue Refunding (CPURR) scheme through extensive simulation experiments where it is shown that OPS and AOPS achieve a much lower expected total travel time and expected total monetary cost for the users compared to the CPURR scheme, without negatively affecting the rest of the network. These results demonstrate the efficiency of personalized (VOT-based) pricing-and-routing schemes.

math.OC↗

arXiv2018

A Model for Partial Kantian Cooperation

In several game situations, the behavior of the players may depend not only on individual interests, but also on what each player considers as the correct thing to do. This work presents a game theoretic model, aiming to describe game situations in which the players' behavior is affected by ethical considerations. Particularly, we assume that they partially follow, Kant's `Categorical Imperative'. The model is stated for games with a continuum of players. The basic assumption made is that the participants perceive that they belong to virtual (imagined) groups, in which they optimize their actions as if they were bound to follow the same strategy. A partially cooperative equilibrium, called $r$-Kant-Nash equilibrium is then introduced. We then study the relationship of the $r$-Kant-Nash equilibrium with the Nash, (Bentham-) Harsanyi, Rawls difference and Roemer solutions. For the case where the set of possible player types is finite, we prove sufficient conditions for the existence and uniqueness of the $r$-Kant-Nash equilibrium and the equilibrium is characterized in terms of a variational inequality. For the case of continuous types, necessary conditions characterizing the partial Kantian equilibria are derived using a reduction to a set of optimal control problems. Finally, some numerical examples are given.

cs.GT↗

arXiv2017

Stochastic Stability in Max-Product and Max-Plus Systems with Markovian Jumps

We study Max-Product and Max-Plus Systems with Markovian Jumps and focus on stochastic stability problems. At first, a Lyapunov function is derived for the asymptotically stable deterministic Max-Product Systems. This Lyapunov function is then adjusted to derive sufficient conditions for the stochastic stability of Max-Product systems with Markovian Jumps. Many step Lyapunov functions are then used to derive necessary and sufficient conditions for stochastic stability. The results for the Max-Product systems are then applied to Max-Plus systems with Markovian Jumps, using an isomorphism and almost sure bounds for the asymptotic behavior of the state are obtained. A numerical example illustrating the application of the stability results on a production system is also given.

eess.SY↗