arXiv subjects

Giuseppe Vizzari

Publications and source records attributed to Giuseppe Vizzari.

At least 19 recordsLinked to original sources

arXiv2025

ConceptBot: Enhancing Robot's Autonomy through Task Decomposition with Large Language Models and Knowledge Graph

ConceptBot is a modular robotic planning framework that combines Large Language Models and Knowledge Graphs to generate feasible and risk-aware plans despite ambiguities in natural language instructions and correctly analyzing the objects present in the environment - challenges that typically arise from a lack of commonsense reasoning. To do that, ConceptBot integrates (i) an Object Property Extraction (OPE) module that enriches scene understanding with semantic concepts from ConceptNet, (ii) a User Request Processing (URP) module that disambiguates and structures instructions, and (iii) a Planner that generates context-aware, feasible pick-and-place policies. In comparative evaluations against Google SayCan, ConceptBot achieved 100% success on explicit tasks, maintained 87% accuracy on implicit tasks (versus 31% for SayCan), reached 76% on risk-aware tasks (versus 15%), and outperformed SayCan in application-specific scenarios, including material classification (70% vs. 20%) and toxicity detection (86% vs. 36%). On SafeAgentBench, ConceptBot achieved an overall score of 80% (versus 46% for the next-best baseline). These results, validated in both simulation and laboratory experiments, demonstrate ConceptBot's ability to generalize without domain-specific training and to significantly improve the reliability of robotic policies in unstructured environments. Website: https://sites.google.com/view/conceptbot

cs.RO↗

arXiv2025

Combining Deep Architectures for Information Gain estimation and Reinforcement Learning for multiagent field exploration

Precision agriculture requires efficient autonomous systems for crop monitoring, where agents must explore large-scale environments while minimizing resource consumption. This work addresses the problem as an active exploration task in a grid environment representing an agricultural field. Each cell may contain targets (e.g., damaged crops) observable from nine predefined points of view (POVs). Agents must infer the number of targets per cell using partial, sequential observations. We propose a two-stage deep learning framework. A pre-trained LSTM serves as a belief model, updating a probabilistic map of the environment and its associated entropy, which defines the expected information gain (IG). This allows agents to prioritize informative regions. A key contribution is the inclusion of a POV visibility mask in the input, preserving the Markov property under partial observability and avoiding revisits to already explored views. Three agent architectures were compared: an untrained IG-based agent selecting actions to maximize entropy reduction; a DQN agent using CNNs over local 3x3 inputs with belief, entropy, and POV mask; and a Double-CNN DQN agent with wider spatial context. Simulations on 20x20 maps showed that the untrained agent performs well despite its simplicity. The DQN agent matches this performance when the POV mask is included, while the Double-CNN agent consistently achieves superior exploration efficiency, especially in larger environments. Results show that uncertainty-aware policies leveraging entropy, belief states, and visibility tracking lead to robust and scalable exploration. Future work includes curriculum learning, multi-agent cooperation with shared rewards, transformer-based models, and intrinsic motivation mechanisms to further enhance learning efficiency and policy generalization.

cs.LG↗

arXiv2025

Why am I seeing this? Towards recognizing social media recommender systems with missing recommendations

Social media plays a crucial role in shaping society, often amplifying polarization and spreading misinformation. These effects stem from complex dynamics involving user interactions, individual traits, and recommender algorithms driving content selection. Recommender systems, which significantly shape the content users see and decisions they make, offer an opportunity for intervention and regulation. However, assessing their impact is challenging due to algorithmic opacity and limited data availability. To effectively model user decision-making, it is crucial to recognize the recommender system adopted by the platform. This work introduces a method for Automatic Recommender Recognition using Graph Neural Networks (GNNs), based solely on network structure and observed behavior. To infer the hidden recommender, we first train a Recommender Neutral User model (RNU) using a GNN and an adapted hindsight academic network recommender, aiming to reduce reliance on the actual recommender in the data. We then generate several Recommender Hypothesis-specific Synthetic Datasets (RHSD) by combining the RNU with different known recommenders, producing ground truths for testing. Finally, we train Recommender Hypothesis-specific User models (RHU) under various hypotheses and compare each candidate with the original used to generate the RHSD. Our approach enables accurate detection of hidden recommenders and their influence on user behavior. Unlike audit-based methods, it captures system behavior directly, without ad hoc experiments that often fail to reflect real platforms. This study provides insights into how recommenders shape behavior, aiding efforts to reduce polarization and misinformation.

cs.IR↗

arXiv2025

SCOOP: A Framework for Proactive Collaboration and Social Continual Learning through Natural Language Interaction andCausal Reasoning

Multimodal information-gathering settings, where users collaborate with AI in dynamic environments, are increasingly common. These involve complex processes with textual and multimodal interactions, often requiring additional structural information via cost-incurring requests. AI helpers lack access to users' true goals, beliefs, and preferences and struggle to integrate diverse information effectively. We propose a social continual learning framework for causal knowledge acquisition and collaborative decision-making. It focuses on autonomous agents learning through dialogues, question-asking, and interaction in open, partially observable environments. A key component is a natural language oracle that answers the agent's queries about environmental mechanisms and states, refining causal understanding while balancing exploration or learning, and exploitation or knowledge use. Evaluation tasks inspired by developmental psychology emphasize causal reasoning and question-asking skills. They complement benchmarks by assessing the agent's ability to identify knowledge gaps, generate meaningful queries, and incrementally update reasoning. The framework also evaluates how knowledge acquisition costs are amortized across tasks within the same environment. We propose two architectures: 1) a system combining Large Language Models (LLMs) with the ReAct framework and question-generation, and 2) an advanced system with a causal world model, symbolic, graph-based, or subsymbolic, for reasoning and decision-making. The latter builds a causal knowledge graph for efficient inference and adaptability under constraints. Challenges include integrating causal reasoning into ReAct and optimizing exploration and question-asking in error-prone scenarios. Beyond applications, this framework models developmental processes combining causal reasoning, question generation, and social learning.

cs.MA↗

arXiv2024

Self-supervised learning for radio-astronomy source classification: a benchmark

The upcoming Square Kilometer Array (SKA) telescope marks a significant step forward in radio astronomy, presenting new opportunities and challenges for data analysis. Traditional visual models pretrained on optical photography images may not perform optimally on radio interferometry images, which have distinct visual characteristics. Self-Supervised Learning (SSL) offers a promising approach to address this issue, leveraging the abundant unlabeled data in radio astronomy to train neural networks that learn useful representations from radio images. This study explores the application of SSL to radio astronomy, comparing the performance of SSL-trained models with that of traditional models pretrained on natural images, evaluating the importance of data curation for SSL, and assessing the potential benefits of self-supervision to different domain-specific radio astronomy datasets. Our results indicate that, SSL-trained models achieve significant improvements over the baseline in several downstream tasks, especially in the linear evaluation setting; when the entire backbone is fine-tuned, the benefits of SSL are less evident but still outperform pretraining. These findings suggest that SSL can play a valuable role in efficiently enhancing the analysis of radio astronomical data. The trained models and code is available at: \url{https://github.com/dr4thmos/solo-learn-radio}

astro-ph.IM↗

arXiv2024

Generalizability analysis of deep learning predictions of human brain responses to augmented and semantically novel visual stimuli

The purpose of this work is to investigate the soundness and utility of a neural network-based approach as a framework for exploring the impact of image enhancement techniques on visual cortex activation. In a preliminary study, we prepare a set of state-of-the-art brain encoding models, selected among the top 10 methods that participated in The Algonauts Project 2023 Challenge [16]. We analyze their ability to make valid predictions about the effects of various image enhancement techniques on neural responses. Given the impossibility of acquiring the actual data due to the high costs associated with brain imaging procedures, our investigation builds up on a series of experiments. Specifically, we analyze the ability of brain encoders to estimate the cerebral reaction to various augmentations by evaluating the response to augmentations targeting objects (i.e., faces and words) with known impact on specific areas. Moreover, we study the predicted activation in response to objects unseen during training, exploring the impact of semantically out-of-distribution stimuli. We provide relevant evidence for the generalization ability of the models forming the proposed framework, which appears to be promising for the identification of the optimal visual augmentation filter for a given task, model-driven design strategies as well as for AR and VR applications.

cs.CV↗

arXiv2023

Exploration and Comparison of Deep Learning Architectures to Predict Brain Response to Realistic Pictures

We present an exploration of machine learning architectures for predicting brain responses to realistic images on occasion of the Algonauts Challenge 2023. Our research involved extensive experimentation with various pretrained models. Initially, we employed simpler models to predict brain activity but gradually introduced more complex architectures utilizing available data and embeddings generated by large-scale pre-trained models. We encountered typical difficulties related to machine learning problems, e.g. regularization and overfitting, as well as issues specific to the challenge, such as difficulty in combining multiple input encodings, as well as the high dimensionality, unclear structure, and noisy nature of the output. To overcome these issues we tested single edge 3D position-based, multi-region of interest (ROI) and hemisphere predictor models, but we found that employing multiple simple models, each dedicated to a ROI in each hemisphere of the brain of each subject, yielded the best results - a single fully connected linear layer with image embeddings generated by CLIP as input. While we surpassed the challenge baseline, our results fell short of establishing a robust association with the data.

q-bio.NC↗

arXiv2023

Radio source analysis services for the SKA and precursors

New developments in data processing and visualization are being made in preparation for upcoming radioastronomical surveys planned with the Square Kilometre Array (SKA) and its precursors. A major goal is enabling extraction of science information from the data in a mostly automated way, possibly exploiting the capabilities offered by modern computing infrastructures and technologies. In this context, the integration of source analysis algorithms into data visualization tools is expected to significantly improve and speed up the cataloguing process of large area surveys. To this aim, the CIRASA (Collaborative and Integrated platform for Radio Astronomical Source Analysis) project was recently started to develop and integrate a set of services for source extraction, classification and analysis into the ViaLactea visual analytic platform and knowledge base archive. In this contribution, we will present the project objectives and tools that have been developed, interfaced and deployed so far on the prototype European Open Science Cloud (EOSC) infrastructure provided by the H2020 NEANIAS project.

astro-ph.IM↗

arXiv2021

Novel EOSC Services for Space Challenges: The NEANIAS First Outcomes

The European Open Science Cloud (EOSC) initiative faces the challenge of developing an agile, fit-for-purpose, and sustainable service-oriented platform that can address the evolving needs of scientific communities. The NEANIAS project plays an active role in the materialization of the EOSC ecosystem by actively contributing to the technological, procedural, strategic and business development of EOSC. We present the first outcomes of the NEANIAS activities relating to co-design and delivery of new innovative services for space research for data management and visualization (SPACE-VIS), map making and mosaicing (SPACE-MOS) and pattern and structure detection (SPACE-ML). We include a summary of collected user requirements driving our services and methodology for their delivery, together with service access details and pointers to future works.

astro-ph.IM↗

arXiv2019

Calibrating Wayfinding Decisions in Pedestrian Simulation Models: The Entropy Map

This paper presents entropy maps, an approach to describing and visualising uncertainty among alternative potential movement intentions in pedestrian simulation models. In particular, entropy maps show the instantaneous level of randomness in decisions of a pedestrian agent situated in a specific point of the simulated environment with an heatmap approach. Experimental results highlighting the relevance of this tool supporting modelers are provided and discussed.

cs.HC↗

arXiv2017

Mining the Social Media Data for a Bottom-Up Evaluation of Walkability

Urbanization represents a huge opportunity for computer applications enabling cities to be managed more efficiently while, at the same time, improving the life quality of their citizens. One of the potential application of this kind of systems is a bottom-up evaluation of the level of walkability of the city (namely the level of usefulness, comfort, safety and attractiveness of an urban area for walking). This is based on the usage of data from social media for the computation of structured indicators describing the actual usage of areas by pedestrians. This paper will present an experimentation of analysis of data about the city of Milano (Italy) acquired from Flickr and Foursquare. The over 500 thousand points, which represent the photos and the POIs collected from the above mentioned social meda, were clustered through an iterative approach based on the DBSCAN algorithm, in order to achieve homogeneous areas defined by the actual activity of inhabitants and tourists rather than by a top down administrative procedure and to supply useful indications on the level of walkability of the city of Milan.

cs.CY↗

arXiv2017

The Role of Compliance in Heterogeneous Interacting Agents: Data from Observations

The dynamics of agent-based systems provide a framework to face the complexity of pedestrian-vehicle interactions in future cities, in which the compliance to traffic norms plays a fundamental role. The data of an observation performed at a non-signalized intersection are presented to provide useful insights for supporting the future development of agent-based models. Results focus on drivers' compliance to crossing pedestrians, describing potentially conflictual interactions among heterogeneous agents. The discussion closes with the potential applications of the collected data set for modelling the phenomenon.

cs.MA↗

arXiv2017

Micro and Macro Pedestrian Dynamics in Counterflow: the Impact of Social Groups

Although it is widely recognised that the presence of groups influences microscopic and aggregated pedestrian dynamics, a precise characterisation of the phenomenon still calls for evidences and insights. The present paper describes micro and macro level original analyses on data characterising pedestrian behaviour in presence of counter-flows and grouping, in particular dyads, acquired through controlled experiments. Results suggest that the presence of dyads and their tendency to walk in a line-abreast formation influences the formation of lanes and, in turn, aggregated observables, such as overall specific flow.

cs.MA↗

arXiv2017

Investigating the effect of social groups in uni-directional pedestrian flow

The influence of cohesion among members of dyads is investigated in scenarios characterized by uni-directional flow by means of a discrete model: a corridor and the egress from a room with a bottleneck of varying width are simulated. The model manages the dynamics of simulated group members with an adaptive mechanism, balancing the probability of movement according to the dispersion of the group; the cohesion mechanism is calibrated through the parameters $κ_c$ and $δ$. All scenarios are simulated with two procedures: (Proc. 1) population composed of individual pedestrians, in order to validate the simulation model and to provide baseline data; (Proc. 2) population including dyads (50% of the simulated pedestrians), in order to verify their impact. In the corridor scenario, the presence of dyads causes a reduction of the velocities and specific flow at medium-high densities. Egress from a square room with a unique central exit produces results in line with recent studies in the literature, but also shows that the dyads negatively affect the dynamics, leading generally to a slower walking speed and a lower pedestrian flow. Ignoring the presence of dyads would lead to an overestimation of egress flows.

cs.MA↗

arXiv2017

Crossing Behaviour of Social Groups: Insights from Observations at Non-signalized Intersection

Environmental, demographical and psychological factors have a demonstrated impact on risky crossing behaviour. In this work we focus on the potential influence of social factors on the considered phenomenon (i.e. group crossing decision). We present the results of a video-recorded observation about the crossing behaviour of singles and dyads at non-signalized intersections. Results showed that crossing behaviour is characterized by three distinct phases: (i) approaching, (ii) appraising (decision making) and (iii) crossing. Dyads walk slower than single pedestrians in all phases. The crossing behaviour of dyads is characterized by the emergence of a leader who takes the decision to cross first, followed by the companion. However, there is no difference between the accepted safety gap of singles and dyads. Understanding factors influencing the crossing decision of social groups represents an important result supporting the development of agent-based simulations of pedestrian-vehicle interactions.

cs.MA↗

arXiv2017

Cellular Automaton Based Simulation of Large Pedestrian Facilities - A Case Study on the Staten Island Ferry Terminals

Current metropolises largely depend on a functioning transport infrastructure and the increasing demand can only be satisfied by a well organized mass transit. One example for a crucial mass transit system is New York City's Staten Island Ferry, connecting the two boroughs of Staten Island and Manhattan with a regular passenger service. Today's demand already exceeds 2500 passengers for a single cycle during peek hours, and future projections suggest that it will further increase. One way to appraise how the system will cope with future demand is by simulation. This contribution proposes an integrated simulation approach to evaluate the system performance with respect to future demand. The simulation relies on a multiscale modeling approach where the terminal buildings are simulated by a microscopic and quantitatively valid cellular automata (CA) and the journeys of the ferries themselves are modeled by a mesoscopic queue simulation approach. Based on the simulation results recommendations with respect to the future demand are given.

cs.MA↗

arXiv2016

Towards Modelling Pedestrian-Vehicle Interactions: Empirical Study on Urban Unsignalized Intersection

The modelling and simulation of the interaction among vehicles and pedestrians during cross-walking is an open challenge for both research and practical computational solutions supporting urban/traffic decision makers and managers. The social cost of pedestrians' risky behaviour pushes the development of a new generation of computational models integrating analytical knowledge, data and experience about the complex dynamics occurring in pedestrian/vehicle interactions, which are not completely understood despite recent efforts. This paper presents the results of a significant data gathering campaign realised at an unsignalized zebra crossing. The selected area of the city of Milan (Italy) is characterised by a significant presence of elderly inhabitants and pedestrian-vehicle risky interactions, testified by a high number of accidents involving pedestrians in the past years. The results concern the analysis of: (i) vehicular and pedestrian traffic volumes; (ii) level of service; (iii) pedestrian-vehicle interactions, considering the impact of ageing on crossing behaviour. Results showed that the phenomenon is characterised by three main phases: approaching, appraising (evaluation of the distance and speed of oncoming vehicles) and crossing. The final objective of the research is to support the development of a microscopic agent-based tool for simulating pedestrian behaviour at unsignalized crosswalks, focusing on the specific needs of the elderly pedestrians.

cs.MA↗

arXiv2016

Avoid or Follow? Modelling Route Choice Based on Experimental Empirical Evidences

Computer-based simulation of pedestrian dynamics reached meaningful results in the last decade, thanks to empirical evidences and acquired knowledge fitting fundamental diagram constraints and space utilization. Moreover, computational models for pedestrian wayfinding often neglect extensive empirical evidences supporting the calibration and validation phase of simulations. The paper presents the results of a set of controlled experiments (with human volunteers) designed and performed to understand pedestrian's route choice. The setting offers alternative paths to final destinations, at different crowding conditions. Results show that the length of paths and level of congestion influence decisions (negative feedback), as well as imitative behaviour of "emergent leaders" choosing a new path (positive feedback). A novel here illustrated model for the simulation of pedestrian route choice captures such evidences, encompassing both the tendency to avoid congestion and to follow emerging leaders. The found conflicting tendencies are modelled with the introduction of a utility function allowing a consistent calibration over the achieved results. A demonstration of the simulated dynamics on a larger scenario will be also illustrated in the paper.

cs.MA↗