arXiv subjects

Gabriel Pena

Publications and source records attributed to Gabriel Pena.

4 recordsLinked to original sources

arXiv2026

On the Structural Limits of Machine Learning Decision Systems: An Information-Theoretic, Interaction-Based, and Stochastic-Dynamical Perspective

Machine learning procedures are commonly evaluated in terms of predictive accuracy and computational efficiency. However, their achievable performance is fundamentally constrained by structural properties of the underlying data-generating process, which are formalized in terms of informational bounds. In this work we examine intrinsic limits of data-driven decision systems from an information-theoretic and interaction-based perspective. We analyze minimal achievable error in classification through Fano-type bounds and precision limits in parametric estimation via the Cram\'er-Rao inequality, emphasizing that such limits depend on the underlying model rather than on algorithmic sophistication alone. We further discuss how implicit assumptions, such as independence, ergodicity, and distributional stability, affect the validity of inferential procedures. Building on interaction-based modeling principles, we review typical frameworks such as Markov Random Fields and potential based representations for encoding dependence mechanisms. We also describe decision systems, including LLM-integrated agent architectures, as feedback-driven stochastic processes where state-dependent dynamics may induce emergent macroscopic behavior. This perspective highlights the importance of having adequate models for the data as a prerequi- site for expanding predictive capability, and situates algorithmic learning within the informational limits imposed by the models.

math.ST↗

arXiv2025

A non-homogeneous Markov early epidemic growth dynamics model. Application to the SARS-CoV-2 pandemic

This work introduces a new markovian stochastic model that can be described as a non-homogeneous Pure Birth process. We propose a functional form of birth rate that depends on the number of individuals in the population and on the elapsed time, allowing us to model a contagion effect. Thus, we model the early stages of an epidemic. The number of individuals then becomes the infectious cases and the birth rate becomes the incidence rate. We obtain this way a process that depends on two competitive phenomena, infection and immunization. Variations in those rates allow us to monitor how effective the actions taken by government and health organizations are. From our model, three useful indicators for the epidemic evolution over time are obtained: the immunization rate, the infection/immunization ratio and the mean time between infections (MTBI). The proposed model allows either positive or negative concavities for the mean value curve, provided the infection/immunization ratio is either greater or less than one. We apply this model to the present SARS-CoV-2 pandemic still in its early growth stage in Latin American countries. As it is shown, the model accomplishes a good fit for the real number of both positive cases and deaths. We analyze the evolution of the three indicators for several countries and perform a comparative study between them. Important conclusions are obtained from this analysis.

q-bio.PE↗

arXiv2025

A non-homogeneous, non-stationary and path-dependent Markov anomalous diffusion model

A novel probabilistic framework for modelling anomalous diffusion is presented. The resulting process is Markovian, non-homogeneous, non-stationary, non-ergodic, and state-dependent. The fundamental law governing this process is driven by two opposing forces: one proportional to the current state, representing the intensity of autocorrelation or contagion, and another inversely proportional to the elapsed time, acting as a damping function. The interplay between these forces determines the diffusion regime, characterized by the ratio of their proportionality coefficients. This framework encompasses various regimes, including subdiffusion, Brownian non-Gaussian, superdiffusion, ballistic, and hyperballistic behaviours. The hyperballistic regime emerges when the correlation force dominates over damping, whereas a balance between these mechanisms results in a ballistic regime, which is also stationary. Crucially, non-stationarity is shown to be necessary for regimes other than ballistic. The model's ability to describe hyperballistic phenomena has been demonstrated in applications such as epidemics, software reliability, and network traffic. Furthermore, deviations from Gaussianity are explored and violations of the Central Limit Theorem are highlighted, supported by theoretical analysis and simulations. It will also be shown that the model exhibits a strong autocorrelation structure due to a position dependent jump probability.

math-ph↗

arXiv2022

Measuring COVID-19 spreading speed through the mean time between infections indicator

We propose to use the mean time between infections (MTBI) metric as obtained from a recently introduced non-homogeneous Markov stochastic model. Different types of parameter calibration are performed. We estimate the MTBI using data from different time windows and from the whole stage history and compare the results. In order to detect waves and stages in the input data, a preprocessing filtering technique is applied. The results of applying this indicator to the COVID-19 reported data of infections from Argentina, Germany and the United States are shown. We find that the MTBI behaves similarly with respect to the different data inputs, whereas the model parameters completely change their behaviour. Evolution over time of the parameters and the MTBI indicator is also shown. We show evidence to support the claim that the MTBI is a rather good indicator in order to measure the spreading speed of an epidemic, having similar values whatever the input data size.

q-bio.PE↗