arXiv subjects

Marco Molinari

Publications and source records attributed to Marco Molinari.

At least 19 recordsLinked to original sources

arXiv2026

Sustainable Heating with Karma: A Simulation Study of the KTH Live-In Lab

Space heating in buildings accounts for 10% of the global CO2 footprint. The widespread adoption of energy-efficient heating technology, e.g., heat pumps, could help reduce this figure, but technology alone may not suffice to reach carbon neutrality. Additionally, human occupants have an important role to play by adopting sustainable heating behaviors, e.g., avoid excessive window opening in the winter or (pre-)heat their units while clean energy is abundant. Thus far demand response policies aimed at promoting these behaviors have been monetary, which discriminates against low-income households and exposes human occupants who do not actively engage with real-time control signals to financial risks. This paper instead investigates the suitability of a non-monetary karma economy for promoting sustainable heating behaviors. Karma leverages the repeated and dynamic nature of heating energy allocations to attain climate targets both fairly and efficiently over time without resorting to financial means. As a first step towards experimentally validating the karma concept with real human occupants in the KTH Live-In Lab, we perform a simulation study on a digital model of the Live-In Lab. The study provides initial estimates of expected effects to guide the design of human-in-the-loop experiments, as well as assists with designing and tuning the karma economy in this context. As a specific example, we investigate how incorporating consumption memory in the form of karma affects window opening behaviors in comparison to conventional memory-less heating operation.

eess.SY↗

arXiv2026

Generalizability of Learning-based Occupancy Detection in Residential Buildings (extended version)

This paper investigates non-intrusive occupancy detection methods for residential buildings using environmental sensor data from the KTH Live-In Lab in Stockholm, Sweden. Three machine learning approaches, namely, logistic regression (LR), support vector machines (SVM), and long short-term memory (LSTM) network enhanced with an attention mechanism, are evaluated in terms of predictive performance and computational complexity. The analysis considers the trade-off between sensor availability (investment cost) and prediction accuracy in real applications, as well as the models' cross-apartment generalizability. Hyperparameters for both the SVM and LSTM models are optimized using Bayesian optimization. All three models are evaluated on data collected from apartments not used during training, and on data generated from a calibrated digital model of the testbed. Results show that all models achieve comparable performance on the same-apartment test data (accuracy of approximately 0.83, F1 score of approximately 0.86). When assessed on cross-apartment data, the LSTM model demonstrates the strongest generalization capability (accuracy of 0.84, F1 score of 0.85), while LR provides a competitive, low-complexity alternative for applications that do not require cross-apartment generalization.

eess.SY↗

arXiv2025

From artificial to circular intelligence to support the well-being of our habitat

The proliferation of machine learning and artificial intelligence redefines the interaction between the anthropogenic and natural elements of our habitat.The use of monitoring tools, processing facilities and the internet of things supports the assessment of planetary health at any given time through automation. However, these data, natural resources and infrastructure intensive technologies are not neutral on the Earth. As the community of AI practitioners works on the creation of tools with minimal socio-environmental impacts, we contribute to the these efforts by proposing a novel conceptual and procedural framework which we call Circular Intelligence or CIntel. CIntel leverages a bottom-up and community-driven approach to learn from the ability of nature to regenerate and adapt. CIntel incorporates ethical principles in its technical design to preserve the stability of the habitat, while also increasing the well-being of its inhabitants by design.

cs.CY↗

arXiv2025

False Discovery Rate Control via Bayesian Mirror Statistic

Simultaneously performing variable selection and inference in high-dimensional models is an open challenge in statistics and machine learning. The increasing availability of vast amounts of variables requires the adoption of specific statistical procedures to accurately select the most important predictors in a high-dimensional space, while being able to control some form of selection error. In this work we adapt the Mirror Statistic approach to False Discovery Rate (FDR) control into a Bayesian modelling framework. The Mirror Statistic, developed in the classic frequentist statistical framework, is a flexible method to control FDR, which only requires mild model assumptions, but requires two sets of independent regression coefficient estimates, usually obtained after splitting the original dataset. Here we propose to rely on a Bayesian formulation of the model and use the posterior distributions of the coefficients of interest to build the Mirror Statistic and effectively control the FDR without the need to split the data. Moreover, the method is very flexible since it can be used with continuous and discrete outcomes and more complex predictors, such as with mixed models. We keep the approach scalable to high-dimensions by relying on Automatic Differentiation Variational Inference and fully continuous prior choices.

stat.ME↗

arXiv2025

Emergent World Representations in OpenVLA

Vision Language Action models (VLAs) trained with policy-based reinforcement learning (RL) encode complex behaviors without explicitly modeling environmental dynamics. However, it remains unclear whether VLAs implicitly learn world models, a hallmark of model-based RL. We propose an experimental methodology using embedding arithmetic on state representations to probe whether OpenVLA, the current state of the art in VLAs, contains latent knowledge of state transitions. Specifically, we measure the difference between embeddings of sequential environment states and test whether this transition vector is recoverable from intermediate model activations. Using linear and non linear probes trained on activations across layers, we find statistically significant predictive ability on state transitions exceeding baselines (embeddings), indicating that OpenVLA encodes an internal world model (as opposed to the probes learning the state transitions). We investigate the predictive ability of an earlier checkpoint of OpenVLA, and uncover hints that the world model emerges as training progresses. Finally, we outline a pipeline leveraging Sparse Autoencoders (SAEs) to analyze OpenVLA's world model.

cs.LG↗

arXiv2025

Fixed Point Explainability

This paper introduces a formal notion of fixed point explanations, inspired by the "why regress" principle, to assess, through recursive applications, the stability of the interplay between a model and its explainer. Fixed point explanations satisfy properties like minimality, stability, and faithfulness, revealing hidden model behaviours and explanatory weaknesses. We define convergence conditions for several classes of explainers, from feature-based to mechanistic tools like Sparse AutoEncoders, and we report quantitative and qualitative results for several datasets and models, including LLMs such as Llama-3.3-70B.

cs.LG↗

arXiv2024

Interpretable Company Similarity with Sparse Autoencoders

Determining company similarity is a vital task in finance, underpinning risk management, hedging, and portfolio diversification. Practitioners often rely on sector and industry classifications such as SIC and GICS codes to gauge similarity, the former being used by the U.S. Securities and Exchange Commission (SEC), and the latter widely used by the investment community. Since these classifications lack granularity and need regular updating, using clusters of embeddings of company descriptions has been proposed as a potential alternative, but the lack of interpretability in token embeddings poses a significant barrier to adoption in high-stakes contexts. Sparse Autoencoders (SAEs) have shown promise in enhancing the interpretability of Large Language Models (LLMs) by decomposing Large Language Model (LLM) activations into interpretable features. Moreover, SAEs capture an LLM's internal representation of a company description, as opposed to semantic similarity alone, as is the case with embeddings. We apply SAEs to company descriptions, and obtain meaningful clusters of equities. We benchmark SAE features against SIC-codes, Industry codes, and Embeddings. Our results demonstrate that SAE features surpass sector classifications and embeddings in capturing fundamental company characteristics. This is evidenced by their superior performance in correlating logged monthly returns - a proxy for similarity - and generating higher Sharpe ratios in co-integration trading strategies, which underscores deeper fundamental similarities among companies. Finally, we verify the interpretability of our clusters, and demonstrate that sparse features form simple and interpretable explanations for our clusters.

cs.CL↗

arXiv2024

Group-SAE: Efficient Training of Sparse Autoencoders for Large Language Models via Layer Groups

SAEs have recently been employed as a promising unsupervised approach for understanding the representations of layers of Large Language Models (LLMs). However, with the growth in model size and complexity, training SAEs is computationally intensive, as typically one SAE is trained for each model layer. To address such limitation, we propose \textit{Group-SAE}, a novel strategy to train SAEs. Our method considers the similarity of the residual stream representations between contiguous layers to group similar layers and train a single SAE per group. To balance the trade-off between efficiency and performance, we further introduce \textit{AMAD} (Average Maximum Angular Distance), an empirical metric that guides the selection of an optimal number of groups based on representational similarity across layers. Experiments on models from the Pythia family show that our approach significantly accelerates training with minimal impact on reconstruction quality and comparable downstream task performance and interpretability over baseline SAEs trained layer by layer. This method provides an efficient and scalable strategy for training SAEs in modern LLMs.

cs.CL↗

arXiv2024

Robust Data-Driven Tube-Based Zonotopic Predictive Control with Closed-Loop Guarantees

This work proposes a robust data-driven tube-based zonotopic predictive control (TZPC) approach for discrete-time linear systems, designed to ensure stability and recursive feasibility in the presence of bounded noise. The proposed approach consists of two phases. In an initial learning phase, we provide an over-approximation of all models consistent with past input and noisy state data using zonotope properties. Subsequently, in a control phase, we formulate an optimization problem, which by integrating terminal ingredients is proven to be recursively feasible. Moreover, we prove that implementing this data-driven predictive control approach guarantees robust exponential stability of the closed-loop system. The effectiveness and competitive performance of the proposed control strategy, compared to recent data-driven predictive control methods, are illustrated through numerical simulations.

eess.SY↗

arXiv2024

A Computationally Efficient Approach to False Discovery Rate Control and Power Maximisation via Randomisation and Mirror Statistic

Simultaneously performing variable selection and inference in high-dimensional regression models is an open challenge in statistics and machine learning. The increasing availability of vast amounts of variables requires the adoption of specific statistical procedures to accurately select the most important predictors in a high-dimensional space, while controlling the false discovery rate (FDR) associated with the variable selection procedure. In this paper, we propose the joint adoption of the Mirror Statistic approach to FDR control, coupled with outcome randomisation to maximise the statistical power of the variable selection procedure, measured through the true positive rate. Through extensive simulations, we show how our proposed strategy allows us to combine the benefits of the two techniques. The Mirror Statistic is a flexible method to control FDR, which only requires mild model assumptions, but requires two sets of independent regression coefficient estimates, usually obtained after splitting the original dataset. Outcome randomisation is an alternative to data splitting that allows to generate two independent outcomes, which can then be used to estimate the coefficients that go into the construction of the Mirror Statistic. The combination of these two approaches provides increased testing power in a number of scenarios, such as highly correlated covariates and high percentages of active variables. Moreover, it is scalable to very high-dimensional problems, since the algorithm has a low memory footprint and only requires a single run on the full dataset, as opposed to iterative alternatives such as multiple data splitting.

stat.ME↗

arXiv2023

Grain boundary segregation and phase separation in ceria-zirconia from atomistic simulation

Doping is the most common strategy employed in the development of new and improved materials. However, predicting the effects of doping on the atomic-scale structure of a material is often difficult or limited to high-end experimental techniques. Doping can induce phase separation in a material, undermining the material's stability. A further complication is that dopant atoms can segregate to interfaces in a material such as grain boundaries (GBs), with consequences for key macroscopic properties of the material such as its conductivity. Here, we describe a computational methodology based on semi-grand canonical Monte Carlo which can be used to probe these phenomena at the atomic scale for metal oxide solid solutions. The methodology can provide precise predictions of the thermodynamic conditions at which phase separation occurs. It can also provide the segregation patterns exhibited by GBs at given conditions. We apply the methodology to one of the most important catalytic materials, ceria-zirconia. Our calculations reveal an interesting richness in the GB segregation in this system. Most GBs we examined exhibited continuous increases in Zr segregation upon Zr doping, with a concomitant reduction in the formation enthalpies of the GBs. However, a few GBs exhibited no segregation at low temperatures. We also observed evidence of first-order complexion transitions in some GBs.

cond-mat.mtrl-sci↗

arXiv2023

Social interactions for a sustainable lifestyle: The design of an experimental case study

Every day we face numerous lifestyle decisions, some dictated by habits and some more conscious, which may or may not promote sustainable living. Aided by digital technology, sustainable behaviors can diffuse within social groups and inclusive communities. This paper outlines a longitudinal experimental study of social influence in behavioral changes toward sustainability, in the context of smart residential homes. Participants are residing in the housing on campus referred to as KTH Live-In Lab, whose behaviors are observed w.r.t. key lifestyle choices, such as food, resources, mobility, consumption, and environmental citizenship. The focus is on the preparatory phase of the case study and the challenges and limitations encountered during its setup. In particular, this work proposes a definition of sustainability indicators for environmentally significant behaviors, and hypothesizes that, through digitalization of a household into a social network of interacting tenants, sustainable living can be promoted. Preliminary results confirm the feasibility of the proposed experimental methodology.

eess.SY↗

arXiv2023

What influences occupants' behavior in residential buildings: An experimental study on window operation in the KTH Live-In Lab

Window-opening and window-closing behaviors play an important role in indoor environmental conditions and therefore have an impact on building energy efficiency. On the other hand, the same environmental conditions drive occupants to interact with windows. Understanding this mutual relationship of interaction between occupants and the residential building is thus crucial to improve energy efficiency without disregarding occupants' comfort. This paper investigates the influence of physical environmental variables (i.e., indoor and outside climate parameters) and categorical variables (i.e., time of the day) on occupants' behavior patterns related to window operation, utilizing a multivariate logistic regression analysis. The data considered in this study are collected during winter months, when the effect on the energy consumption of the window operation is the highest, at a Swedish residential building, the KTH Live-In Lab, accommodating four occupants in separate studio apartments. Although all the occupants seem to share a sensitivity to some common factors, such as air quality and time of the day, we can also observe individual variability with respect to the most significant drivers influencing window operation behaviors.

eess.SY↗

arXiv2023

Closing the gap between research and projects in climate change innovation in Europe

Innovation is a key component to equip our society with tools to adapt to new climatic conditions. The development of research-action interfaces shifts useful ideas into operationalized knowledge allowing innovation to flourish. In this paper we quantify the existing gap between climate research and innovation action in Europe using a novel framework that combines artificial intelligence (AI) methods and network science. We compute the distance between key topics of research interest from peer review publications and core issues tackled by innovation projects funded by the most recent European framework programmes. Our findings reveal significant differences exist between and within the two layers. Economic incentives, agricultural and industrial processes are differently connected to adaptation and mitigation priorities. We also find a loose research-action connection in bioproducts, biotechnologies and risk assessment practices, where applications are still too few compared to the research insights. Our analysis supports policy-makers to measure and track how research funding result in innovation action, and to adjust decisions if stated priorities are not achieved.

cs.CY↗

arXiv2022

Robust Data-Driven Predictive Control of Unknown Nonlinear Systems using Reachability Analysis

This work proposes a robust data-driven predictive control approach for unknown nonlinear systems in the presence of bounded process and measurement noise. Data-driven reachable sets are employed for the controller design instead of using an explicit nonlinear system model. Although the process and measurement noise are bounded, the statistical properties of the noise are not required to be known. By using the past noisy input-output data in the learning phase, we propose a novel method to over-approximate reachable sets of an unknown nonlinear system. Then, we propose a data-driven predictive control approach to compute safe and robust control policies from noisy online data. The constraints are guaranteed in the control phase with robust safety margins through the effective use of the predicted output reachable set obtained in the learning phase. Finally, a numerical example validates the efficacy of the proposed approach and demonstrates comparable performance with a model-based predictive control approach.

eess.SY↗

arXiv2022

Bayesian Dynamic Network Modelling: an application to metabolic associations in cardiovascular diseases

We propose a novel approach to the estimation of multiple Graphical Models to analyse temporal patterns of association among a set of metabolites over different groups of patients. Our motivating application is the Southall And Brent REvisited (SABRE) study, a tri-ethnic cohort study conducted in the UK. We are interested in identifying potential ethnic differences in metabolite levels and associations as well as their evolution over time, with the aim of gaining a better understanding of different risk of cardio-metabolic disorders across ethnicities. Within a Bayesian framework, we employ a nodewise regression approach to infer the structure of the graphs, borrowing information across time as well as across ethnicities. The response variables of interest are metabolite levels measured at two time points and for two ethnic groups, Europeans and South-Asians. We use nodewise regression to estimate the high-dimensional precision matrices of the metabolites, imposing sparsity on the regression coefficients through the dynamic horseshoe prior, thus favouring sparser graphs. We provide the code to fit the proposed model using the software Stan, which performs posterior inference using Hamiltonian Monte Carlo sampling, as well as a detailed description of a block Gibbs sampling scheme.

stat.ME↗

arXiv2021

Data-Driven Control and Data-Poisoning attacks in Buildings: the KTH Live-In Lab case study

This work investigates the feasibility of using input-output data-driven control techniques for building control and their susceptibility to data-poisoning techniques. The analysis is performed on a digital replica of the KTH Livein Lab, a non-linear validated model representing one of the KTH Live-in Lab building testbeds. This work is motivated by recent trends showing a surge of interest in using data-based techniques to control cyber-physical systems. We also analyze the susceptibility of these controllers to data-poisoning methods, a particular type of machine learning threat geared towards finding imperceptible attacks that can undermine the performance of the system under consideration. We consider the Virtual Reference Feedback Tuning (VRFT), a popular data-driven control technique, and show its performance on the KTH Live-In Lab digital replica. We then demonstrate how poisoning attacks can be crafted and illustrate the impact of such attacks. Numerical experiments reveal the feasibility of using data-driven control methods for finding efficient control laws. However, a subtle change in the datasets can significantly deteriorate the performance of VRFT.

eess.SY↗

arXiv2020

Maximizing Privacy in MIMO Cyber-Physical Systems Using the Chapman-Robbins Bound

Privacy breaches of cyber-physical systems could expose vulnerabilities to an adversary. Here, privacy leaks of step inputs to linear-time-invariant systems are mitigated through additive Gaussian noise. Fundamental lower bounds on the privacy are derived, which are based on the variance of any estimator that seeks to recreate the input. Fully private inputs are investigated and related to transmission zeros. Thereafter, a method to increase the privacy of optimal step inputs is presented and a privacy-utility trade-off bound is derived. Finally, these results are verified on data from the KTH Live-In Lab Testbed, showing good correspondence with theoretical results.

eess.SY↗