arXiv subjects

Marco D'Alessandro

Publications and source records attributed to Marco D'Alessandro.

17 recordsLinked to original sources

arXiv2026

Dark exciton signatures in the infrared transient absorption of MoS$_2$ monolayer

Dark excitons play a central role in the nonequilibrium dynamics of two-dimensional semiconductors, but remain difficult to characterize. Transient-absorption experiments, with probes tuned in exciton-exciton transitions energy range (exc-tr-abs), can detect excitations from any populated dark excitons, including symmetry-forbidden, spin-forbidden, and finite-momentum ones. In this work, we develop a $GW$+BSE scheme for computing exc-tr-abs spectra from arbitrary populated exciton distributions. These dark excitons are included equally by evaluating exciton-exciton dipoles in a locally smooth gauge, including intra- and inter-band contributions. For monolayer MoS$_2$, the exc-tr-abs signal arises from the $Γ$, $K$, $M$, and $Q$ valleys, differing substantially from the $Γ$-only interpretation. Exciton-exciton dipoles show similar intensities across these valleys, while their weights are dictated by initial excitonic populations. State- and spin-resolved analyses assign the peaks to $1s \rightarrow 2p$ and $1s \rightarrow 3p$ transitions from both spin-flip and spin-conserving A and B excitons across valleys and momenta.

cond-mat.mtrl-sci↗

arXiv2026

Grounded world models in biological organisms and future embodied AI

Recent advances in generative and embodied AI have been driven by large-scale predictive learning over multimodal data. However, the resulting systems remain largely based on passive training regimes where linguistic regularities create the scaffold onto which information from other modalities is attached. Conversely, neuroscience and cognitive science suggest that biological intelligence is organized in the opposite way, where grounded world models acquired through interaction with the environment provide the semantic scaffold to which language is attached. Here, we illustrate five examples of neural circuits supporting grounded world modelling, which underlie navigation in physical and conceptual spaces, affordance-based perception and interaction with objects, active perception and exploratory learning, allostatic control and emotion, and the distinction between self- and world-generated outcomes. These examples highlight several features largely missing from current embodied AI, including the role of intrinsic dynamics as a foundation for learning, the centrality of action in aligning these dynamics with the external world, the prominence of autonomous experience and open-ended learning over passive assimilation of externally provided data, and the fact that early predictive and control mechanisms scaffold higher cognitive abilities such as reasoning, conceptual navigation, planning, imagination, understanding others' minds, and communication. Finally, we discuss whether and how principles derived from biological systems may inform future embodied AI, including training regimes based on social interaction to construct world models that are not only grounded but also socially shared and aligned with human norms and values.

q-bio.NC↗

arXiv2026

Towards terahertz excitons in hydrogenated graphene superlattices

Carbon nanostructures, such as nanotubes and graphene nanoribbons, exhibit unique electronic and optical properties that make them very promising candidates for terahertz components. However, carbon nanotube and nanoribbon monolithic on-chip integration is challenging because it may results in significant change of their intrinsic properties after an embedment into a substrate. We investigate with first principles theoretical methods the successful routes of such integration and calculate electronic and optical properties of the integrated structures -- two-dimensional graphene superlattices, where quasi-metallic and dielectric regions alternate by selective hydrogenation of graphene. It is shown that chemical engineering of the graphene surface can lead to strong and well-isolated excitonic absorption peaks in the far-infrared and possibly even terahertz frequencies.

cond-mat.mes-hall↗

arXiv2026

GENESIS: A Generative Model of Episodic-Semantic Interaction

A central challenge in cognitive neuroscience is to explain how semantic and episodic memory, two major forms of declarative memory, typically associated with cortical and hippocampal processing, interact to support learning, recall, and imagination. Despite significant advances, we still lack a unified computational framework that jointly accounts for core empirical phenomena across both semantic and episodic processing domains. Here, we introduce the Generative Episodic-Semantic Integration System (GENESIS), a computational model that formalizes memory as the interaction between two limited-capacity generative systems: a Cortical-VAE, supporting semantic learning and generalization, and a Hippocampal-VAE, supporting episodic encoding and retrieval within a retrieval-augmented generation (RAG) architecture. GENESIS reproduces hallmark behavioral findings, including generalization in semantic memory, recognition, serial recall effects and gist-based distortions in episodic memory, and constructive episodic simulation, while capturing their dynamic interactions. The model elucidates how capacity constraints shape the fidelity and memorability of experiences, how semantic processing introduces systematic distortions in episodic recall, and how episodic replay can recombine previous experiences. Together, these results provide a principled account of memory as an active, constructive, and resource-bounded process. GENESIS thus advances a unified theoretical framework that bridges semantic and episodic memory, offering new insights into the generative foundations of human cognition.

q-bio.NC↗

arXiv2025

Sensory robustness through top-down feedback and neural stochasticity in recurrent vision models

Biological systems leverage top-down feedback for visual processing, yet most artificial vision models succeed in image classification using purely feedforward or recurrent architectures, calling into question the functional significance of descending cortical pathways. Here, we trained convolutional recurrent neural networks (ConvRNN) on image classification in the presence or absence of top-down feedback projections to elucidate the specific computational contributions of those feedback pathways. We found that ConvRNNs with top-down feedback exhibited remarkable speed-accuracy trade-off and robustness to noise perturbations and adversarial attacks, but only when they were trained with stochastic neural variability, simulated by randomly silencing single units via dropout. By performing detailed analyses to identify the reasons for such benefits, we observed that feedback information substantially shaped the representational geometry of the post-integration layer, combining the bottom-up and top-down streams, and this effect was amplified by dropout. Moreover, feedback signals coupled with dropout optimally constrained network activity onto a low-dimensional manifold and encoded object information more efficiently in out-of-distribution regimes, with top-down information stabilizing the representational dynamics at the population level. Together, these findings uncover a dual mechanism for resilient sensory coding. On the one hand, neural stochasticity prevents unit-level co-adaptation albeit at the cost of more chaotic dynamics. On the other hand, top-down feedback harnesses high-level information to stabilize network activity on compact low-dimensional manifolds.

q-bio.NC↗

arXiv2024

Multimodal Parameter-Efficient Few-Shot Class Incremental Learning

Few-Shot Class Incremental Learning (FSCIL) is a challenging continual learning task, where limited training examples are available during several learning sessions. To succeed in this task, it is necessary to avoid over-fitting new classes caused by biased distributions in the few-shot training sets. The general approach to address this issue involves enhancing the representational capability of a pre-defined backbone architecture by adding special modules for backward compatibility with older classes. However, this approach has not yet solved the dilemma of ensuring high classification accuracy over time while reducing the gap between the performance obtained on larger training sets and the smaller ones. In this work, we propose an alternative approach called Continual Parameter-Efficient CLIP (CPE-CLIP) to reduce the loss of information between different learning sessions. Instead of adapting additional modules to address information loss, we leverage the vast knowledge acquired by CLIP in large-scale pre-training and its effectiveness in generalizing to new concepts. Our approach is multimodal and parameter-efficient, relying on learnable prompts for both the language and vision encoders to enable transfer learning across sessions. We also introduce prompt regularization to improve performance and prevent forgetting. Our experimental results demonstrate that CPE-CLIP significantly improves FSCIL performance compared to state-of-the-art proposals while also drastically reducing the number of learnable parameters and training costs.

cs.CV↗

arXiv2023

Transitions in Xenes between excitonic, topological and trivial insulator phases: influence of screening, band dispersion and external electric field

Using a variational approach, the binding energies $E_b$ of the lowest bound excitons in Xenes under varying electric field are investigated. The internal exciton motion is described both by Dirac electron dispersion and in effective-mass approximation, while the screened electron-hole attraction is modeled by a Rytova-Keldysh potential with a 2D electronic polarizability $α_{2{\rm D}}$. The most important parameters as spin-orbit-induced gap $E_g$, Fermi velocity $v_F$ and $α_{2{\rm D}}$ are taken from ab initio density functional theory calculations. In addition, $α_{2{\rm D}}$ is approximated in two different ways. The relation of $E_b$ and $E_g$ is ruled by the screening. The existence of an excitonic insulator phase with $E_b>E_g$ sensitively depends on the chosen $α_{2{\rm D}}$. The values of $E_g$ and $α_{2{\rm D}}$ are strongly modified by a vertical external electric bias $U$, which defines a transition from the topological into a trivial insulator at $U=E_g/2$, with the exception of plumbene. Within the Dirac approximation, but also within the effective mass description of the kinetic energy, the treatment of screening dominates the appearance or non-appearance of an excitonic insulator phase. Gating does not change the results: the prediction done at zero electric field is confirmed when a vertical electric field is applied. Finally, Many-Body perturbation theory approaches based on the Green's function method, applied to stanene, confirm the absence of an excitonic insulator phase, thus validating our results obtained by ab initio modeling of $α_{2{\rm D}}$.

cond-mat.mtrl-sci↗

arXiv2021

Amortized Bayesian model comparison with evidential deep learning

Comparing competing mathematical models of complex natural processes is a shared goal among many branches of science. The Bayesian probabilistic framework offers a principled way to perform model comparison and extract useful metrics for guiding decisions. However, many interesting models are intractable with standard Bayesian methods, as they lack a closed-form likelihood function or the likelihood is computationally too expensive to evaluate. With this work, we propose a novel method for performing Bayesian model comparison using specialized deep learning architectures. Our method is purely simulation-based and circumvents the step of explicitly fitting all alternative models under consideration to each observed dataset. Moreover, it requires no hand-crafted summary statistics of the data and is designed to amortize the cost of simulation over multiple models and observable datasets. This makes the method particularly effective in scenarios where model fit needs to be assessed for a large number of datasets, so that per-dataset inference is practically infeasible.Finally, we propose a novel way to measure epistemic uncertainty in model comparison problems. We demonstrate the utility of our method on toy examples and simulated data from non-trivial models from cognitive science and single-cell neuroscience. We show that our method achieves excellent results in terms of accuracy, calibration, and efficiency across the examples considered in this work. We argue that our framework can enhance and enrich model-based analysis and inference in many fields dealing with computational models of natural processes. We further argue that the proposed measure of epistemic uncertainty provides a unique proxy to quantify absolute evidence even in a framework which assumes that the true data-generating model is within a finite set of candidate models.

stat.ML↗

arXiv2020

A Bayesian brain model of adaptive behavior: An application to the Wisconsin Card Sorting Task

Adaptive behavior emerges through a dynamic interaction between cognitive agents and changing environmental demands. The investigation of information processing underlying adaptive behavior relies on controlled experimental settings in which individuals are asked to accomplish demanding tasks whereby a hidden state or an abstract rule has to be learned dynamically. Although performance in such tasks is regularly considered as a proxy for measuring high-level cognitive processes, the standard approach consists in summarizing response patterns by simple heuristic scoring measures. With this work, we propose and validate a new computational Bayesian model accounting for individual performance in the established Wisconsin Card Sorting Test. We embed the new model within the mathematical framework of Bayesian Brain Theory, according to which beliefs about the hidden environmental states are dynamically updated following the logic of Bayesian inference. Our computational model maps distinct cognitive processes into separable, neurobiologically plausible, information-theoretic constructs underlying observed response patterns. We assess model identification and expressiveness in accounting for meaningful human performance through extensive simulation studies. We further apply the model to real behavioral data in order to highlight the utility of the proposed model in recovering cognitive dynamics at an individual level. Practical and theoretical implications of our computational modeling approach for clinical and cognitive neuroscience research are finally discussed, as well as potential future improvements.

q-bio.NC↗

arXiv2020

Real-time modelling of Optical orientation in GaAs: generation and decay of the degree of spin polarization

We present a real-time abinitio description of optical orientation in bulk GaAs due to the coupling with an ultrashort circular polarized laser source. The injection of spin polarized electrons in the conduction band is correctly reproduced, and a non vanishing spin polarization ($\mathbf{P}$) parallel to the direction of propagation of the laser ($z$) emerges. A detailed analysis of the generation and the evolution of $\mathbf{P}(t)$ is discussed. The single $\mathbf{k}$-point dynamics is a motion of precession around a fixed axis with constant $|\mathbf{P}|$ and fixed frequency. Instead, the $\mathbf{k}$-integrated signal shows only a time dependent $P_z(t)$ and decays few pico seconds after the end of the laser pump due to decoherence. Decoherence emerges since the individual contributions activated by the pump give rise to destructive interference. We interpret the results in terms of the \emph{free induction decay} mechanism proposed some years ago. For the first time we are able to reproduce such effect in a full abinitio fashion, giving a quantitative estimate of the associated decay time. Our result also shows a possible explanation to the time decay of spin magnetization observed in many real-time abinitio simulations.

cond-mat.mtrl-sci↗

arXiv2020

The BepiColombo SERENA/ELENA Instrument On-Ground Testing with the ELENA Special Check Out Equipment (SCOE)

The neutral particles sensor ELENA (Emitted Low Energy Neutral Atoms) for the ESA/JAXA BepiColombo mission to Mercury (in the SERENA instrument suite) is devoted to measure low energetic neutral atoms. The main goal of the experiment is measuring the sputtering emission from planetary surfaces, from E=20eV up to E=5keV, within 1D (2 deg. x 76 deg. ). ELENA original project had also a particle discrimination system based on Time-of-Flight (TOF) of particles through the shutter on the Micro Channel Plates detector (MCP), it has been withdrawn from the flight model due to design and development problems. The ELENA SCOE is the configuration/testing system of ELENA, it allows to command operations and to set up configuration parameters on the instrument and to monitor the incoming data. The TC/TM simulation/encoding/decoding software is developed respecting the CCSDS/ECSS standards implemented by ESA, and it's SCOS2000 compatible. TC generation, HK data monitoring and basic science data analysis are operated by the SERENA EGSE, developed by the Finnish Meteorological Institute (FMI), Helsinki, Finland. The data stream outcoming from the EGSE is then preprocessed from TM to user readable formats: FITS and then ASCII csv tables with metadata collected in a detached XML file, called label. This task is performed using the PacketLib, ProcessorLib, and DISCoS (PPD) framework and is going to be used as the first level prototype of the BepiColombo Science Ground Segment processing pipeline, based in ESAC, Madrid, Spain and implemented using the PDS4 data format.

astro-ph.IM↗

arXiv2019

A state space approach to dynamic modeling of mouse-tracking data

Mouse-tracking recording techniques are becoming very attractive in experimental psychology. They provide an effective means of enhancing the measurement of some real-time cognitive processes involved in categorization, decision-making, and lexical decision tasks. Mouse-tracking data are commonly analysed using a two-step procedure which first summarizes individuals' hand trajectories with independent measures, and then applies standard statistical models on them. However, this approach can be problematic in many cases. In particular, it does not provide a direct way to capitalize the richness of hand movement variability within a consistent and unified representation. In this article we present a novel, unified framework for mouse-tracking data. Unlike standard approaches to mouse-tracking, our proposal uses stochastic state-space modeling to represent the observed trajectories in terms of both individual movement dynamics and experimental variables. The model is estimated via a Metropolis-Hastings algorithm coupled with a non-linear recursive filter. The characteristics and potentials of the proposed approach are illustrated using a lexical decision case study. The results highlighted how dynamic modeling of mouse-tracking data can considerably improve the analysis of mouse-tracking tasks and the conclusions researchers can draw from them.

stat.AP↗

arXiv2019

Locality and Computational Reliability of Linear Response Calculations for Molecular Systems

By performing a critical analysis of the fundamental equations of linear-response (LR) formalism in molecules, we explore the interplay between locality of the response density operator and numerical convergence of LR-related quantities. We show that for frequencies below the first ionization potential (IP) of the system, it is possible to express the response density by employing localized states only. Above this threshold energy, such a locality property cannot be achieved. Such considerations may be transposed in terms of the molecule's excited states. We show that not all the system's excitations can be considered on equal footing. There is a discrete sector of excitations -- which may also extend above IP -- that can be parametrized by observable, localized states, which can be computationally expressed with high precision, provided an adequate level of completeness. We present indicators that can help to quantify such potential observable properties of an excitation, that can be evaluated in any discretization scheme. The remaining excitation modes belong to a continuum spectrum that, on the contrary, is not directly associated to observable properties and can only be effectively represented in a given computational setup. Such considerations are important not only for reproducibility of the results among different computer codes employing diverse formalisms, but also in view of providing a deeper understanding on the impact of models' approximations on the scientific outcomes of the simulation.

physics.comp-ph↗

arXiv2016

Multiparticle correlation expansion of relative entropy in lattice systems

This paper deals with the construction of the multiparticle correlation expansion of relative entropy for lattice systems. Thanks to this analysis we are able to express the statistical distance between two systems as a series built over clusters of increasing dimension. Each addend is written in terms of correlation functions and expresses the contribution to the relative entropy due to structural information inside the selected cluster. We present a general procedure for the explicit construction of all the terms of the series. As a first application of this result, we show that the coefficients of the multiparticle correlation expansion of the excess entropy can be computed from our formula, as a particular case.

cond-mat.stat-mech↗

arXiv2012

Maximum-entropy Monte Carlo method for the inversion of the structure factor in simple classical systems

We present a method for the evaluation of the interaction potential of an equilibrium classical system starting from the (partial) knowledge of its structure factor. The procedure is divided into two phases both of which are based on the maximum entropy principle of information theory. First we determine the maximum entropy estimate of the radial distribution function constrained by the information contained in the structure factor. Next we invert the pair function and extract the interaction potential. The method is tested on a Lennard-Jones fluid at high density and the reliability of its results with respect to the missing information in the structure factor data are discussed. Finally, it is applied to the experimental data of liquid sodium at 100$^{\circ}$C.

cond-mat.stat-mech↗

arXiv2010

Information-theory-based solution of the inverse problem in classical statistical mechanics

We present a procedure for the determination of the interaction potential from the knowledge of the radial pair distribution function. The method, realized inside an inverse Monte Carlo simulation scheme, is based on the application of the Maximum Entropy Principle of information theory and the interaction potential emerges as the asymptotic expression of the transition probability. Results obtained for high density monoatomic fluids are very satisfactory and provide an accurate extraction of the potential, despite a modest computational effort.

cond-mat.stat-mech↗

arXiv2005

A wide class of four point functions of BPS operators in N=4 SYM at order g^4

The calculation of a large family of four point functions of general BPS operators in N=4 SYM is reduced to the evaluation of colour contractions. For 1/2 BPS operators O_Δthe explicit results at order g^4 for the function < O_n O_2 O_n O_2 > are given up to n=6. The OPE of the general result is performed up to the second order in the short distance expansion parameter. Two examples are given, in which the mixing of the operators in the intermediate channel can be resolved using four point functions computed by this method.

hep-th↗