arXiv subjects

Daniele Marinazzo

Publications and source records attributed to Daniele Marinazzo.

At least 19 recordsLinked to original sources

arXiv2026

Statistical Dark Matter: Synergy is everywhere but is hard to capture

The information-theoretic construct of synergy refers to the statistical structure that is contained in three or more variables, but not in any subset of them. Here we leverage recent advances in multivariate information theory to show that synergy is far more prevalent in complex systems than previously thought. In particular, we show that many-body systems tend to become strongly dominated by synergy as the number of subcomponents grows. At the same time, our results also reveal that commonly used tools for statistical modelling severely underestimate synergistic structures. These findings imply that synergy constitutes a prevalent, yet often invisible, informational component of complex systems. With a certain poetic licence, we interpret these results as suggesting that synergy is the dark matter of statistics - interdependencies that we know exist, but standard instruments fail to detect.

cs.IT↗

arXiv2026

From Statistical to Structural Synergy: A Predictability Framework to Quantify the Effects due to High-Order Mechanisms

High-order interactions are increasingly recognized as a hallmark of collective dynamics in complex systems. The relationship between high-order behaviours (HOBs), observed as synergistic or redundant statistical dependencies, and high-order mechanisms (HOMs), related to the structural or dynamical rules of the data-generating process, remains difficult to establish from data. We introduce a predictability-based framework to disentangle these two levels of description in complex network systems. Structural synergy is defined as the excess predictive power gained when two sources are considered jointly beyond the best additive description and is estimated through polynomial regression by comparing a model with interaction terms against an additive model. Simulations show that dependencies among sources reflecting HOBs can arise even in the absence of HOMs, while synergy due to a non-additive mechanism may remain hidden when the observed synergy-redundancy balance is dominated by redundancy and become detectable only through structural synergy. Applications to climate and source-reconstructed cortical EEG dynamics reveal significant non-additive predictive components despite predominantly redundancy-dominated HOBs. These findings emphasize that HOBs and HOMs can dissociate: a system may display redundancy-dominated HOBs while still containing a significant synergistic mechanism. The proposed framework supports a mechanistically informed interpretation of complex-system dynamics and may help to identify when mechanism-based models are needed to predict the response of a system to perturbations or interventions, while also recognizing that structural synergy should be interpreted as evidence of non-additive predictive structure rather than as a direct identification of the underlying generative mechanism.

stat.ME↗

arXiv2026

Dissecting Spectral Granger Causality through Partial Information Decomposition

Granger causality (GC), a popular statistical method for the inference of directional influences between time series measured from a complex network, is sensitive to high-order (non-pairwise) interactions which fundamentally shape the collective network dynamics. This work introduces Partial Decomposition of Granger Causality (PDGC), a tool eliciting redundant and synergistic causal interactions in the pattern of information flow between the subsystems of physiological networks. The tool exploits the framework of partial information decomposition to dissect the multivariate GC from a set of driver random processes to a target process into unique effects carried exclusively by each driver, redundant effects carried identically by more drivers, and synergistic effects carried jointly by some drivers but not by any of them individually. Computation is based on multivariate state-space models expanded in the frequency domain to assess PDGC both in specific bands of physiological interest and in the time domain after whole-band integration. The validation on benchmark simulations demonstrates that the measures of unique, redundant, and synergistic GC reflect the underlying causal mechanisms and are computationally reliable. The application to arterial pressure, respiration, cerebral blood velocity and heart period variability reveals striking differences in the response to postural stress of patients prone to neurally-mediated syncope compared to healthy controls. The extraction of high-order causality patterns from the spectral GC favors dissecting the mechanisms of causal influence underlying multivariate interactions among oscillatory processes in many data-driven applications of network science.

stat.ME↗

arXiv2026

Cohort-amortized personalization: navigating the privacy-utility frontier for virtual brain twins

Personalized generative brain models require individual neuroimaging data that privacy constraints and re-identification risk make difficult to share, while per-subject fitting procedures cost hours of compute -- limiting clinical translation and multi-site collaboration. We introduce cohort-amortized personalization (CAP), which replaces data sharing with model sharing: a neural density estimator is trained on simulations from a mechanistic whole-brain model under a low-rank cohort prior, and only the compact estimator is distributed, so new subjects are personalized in seconds on their own data alone. To make this prior both compact and atlas-independent, a cross-atlas autoencoder (CrossCoder) maps connectomes from 20 anatomical atlases into a shared latent space, enabling deployment across sites with heterogeneous atlases. We validate CAP on two cohorts: 21 patients with drug-resistant epilepsy (epileptogenic-zone localization F1=0.56) and 832 subjects from the 1000BRAINS aging cohort (predicted age r=0.44); in both, CAP matches or exceeds per-subject inference with hours-to-seconds speed-up. Because the shared artifact couples a cohort prior to a mechanistic simulator, it can serve as a mechanistic surrogate supporting in-silico experimentation and synthetic-cohort generation without raw-data access -- a governance-audited alternative we term synthetic access, allowing for wider adoption of personalized modeling in more diverse settings.

q-bio.NC↗

arXiv2025

Partial Information Rate Decomposition

Partial Information Decomposition (PID) is a principled and flexible method to unveil complex high-order interactions in multi-unit network systems. Though being defined exclusively for random variables, PID is ubiquitously applied to multivariate time series taken as realizations of random processes with temporal statistical structure. Here, to overcome the incorrect depiction of high-order effects by PID schemes applied to dynamic networks, we introduce the framework of Partial Information Rate Decomposition (PIRD). PIRD is first formalized applying lattice theory to decompose the information shared dynamically between a target random process and a set of source processes, and then implemented for Gaussian processes through a spectral expansion of information rates. The new framework is validated in simulated network systems and demonstrated in the practical analysis of time series from large-scale climate oscillations.

stat.ME↗

arXiv2025

Decomposing Multivariate Information Rates in Networks of Random Processes

The Partial Information Decomposition (PID) framework has emerged as a powerful tool for analyzing high-order interdependencies in complex network systems. However, its application to dynamic processes remains challenging due to the implicit assumption of memorylessness, which often falls in real-world scenarios. In this work, we introduce the framework of Partial Information Rate Decomposition (PIRD) that extends PID to random processes with temporal correlations. By leveraging mutual information rate (MIR) instead of mutual information (MI), our approach decomposes the dynamic information shared by multivariate random processes into unique, redundant, and synergistic contributions obtained aggregating information rate atoms in a principled manner. To solve PIRD, we define a pointwise redundancy rate function based on the minimum MI principle applied locally in the frequency-domain representation of the processes. The framework is validated in benchmark simulations of Gaussian systems, demonstrating its advantages over traditional PID in capturing temporal correlations and showing how the spectral representation may reveal scale-specific higher-order interactions that are obscured in the time domain. Furthermore, we apply PIRD to a physiological network comprising cerebrovascular and cardiovascular variables, revealing frequency-dependent redundant information exchange during a protocol of postural stress. Our results highlight the necessity of accounting for the full temporal statistical structure and spectral content of vector random processes to meaningfully perform information decomposition in network systems with dynamic behavior such as those typically encountered in neuroscience and physiology.

stat.ME↗

arXiv2025

Connectome brain fingerprinting: terminology, measures, and target properties

Distinguishing one person from another (what biometricians call recognition) is extremely relevant for different aspects of life. Traditional biometric modalities (fingerprint, face, iris, voice) rely on unique, stable features that reliably differentiate individuals. Recently, the term fingerprinting has gained popularity in neuroscience, with a growing number of studies adopting the term to describe various brain based metrics derived from different techniques. However, we think there is a mismatch between its widely accepted meaning in the biometric community and some brain based metrics. Many of these measures do not satisfy the strict definition of a biometric fingerprint that is, a stable trait that uniquely identifies an individual. In this study we discuss some issues that may generate confusion in this context and suggest how to treat the question in the future. In particular, we review how fingerprint is currently used in the neuroscience literature, highlight mismatches with the biometric community definition, and offer clear guidelines for distinguishing genuine biometric fingerprints from exploratory similarity metrics. By clarifying terminology and criteria, we aim to align practices and facilitate communication across fields.

q-bio.QM↗

arXiv2025

Localizing synergies of hidden factors across complex systems: resting brain networks and HeLa gene expression profile as case studies

Factor analysis is a well-known statistical method to describe the variability of observed variables in terms of a smaller number of unobserved latent variables called factors. Even though latent factors are conceptually independent of each other, their influence on the observed variables is often joint and synergistic. We propose to quantify the synergy of the joint influence of factors on the observed variables using the O-information, a recently introduced metrics to assess high order dependencies in complex systems, in a new framework where latent factors and observed variables are jointly analyzed in terms of their joint informational character. Two case studies are reported: analyzing resting fMRI data, we find that DMN and FP networks show the highest synergy, consistently with their crucial role in higher cognitive functions; concerning HeLa cells, we find that the most synergistic gene is STK-12 (AURKB), suggesting that this gene is involved in controlling the HeLa cell cycle. We believe that this approach, representing a bridge between factor analysis and the field of high-order interactions, will find wide application across several domains.

q-bio.QM↗

arXiv2025

Assessing high-order effects in feature importance via predictability decomposition

Leveraging the large body of work devoted in recent years to describe redundancy and synergy in multivariate interactions among random variables, we propose a novel approach to quantify cooperative effects in feature importance, one of the most used techniques for explainable artificial intelligence. In particular, we propose an adaptive version of a well-known metric of feature importance, named Leave One Covariate Out (LOCO), to disentangle high-order effects involving a given input feature in regression problems. LOCO is the reduction of the prediction error when the feature under consideration is added to the set of all the features used for regression. Instead of calculating the LOCO using all the features at hand, as in its standard version, our method searches for the multiplet of features that maximize LOCO and for the one that minimize it. This provides a decomposition of the LOCO as the sum of a two-body component and higher-order components (redundant and synergistic), also highlighting the features that contribute to building these high-order effects alongside the driving feature. We report the application to proton/pion discrimination from simulated detector measures by GEANT.

physics.data-an↗

arXiv2025

Cognitive networks highlight differences and similarities in the STEM mindsets of human and LLM-simulated trainees, experts and academics

Understanding attitudes towards STEM means quantifying the cognitive and emotional ways in which individuals, and potentially large language models too, conceptualise such subjects. This study uses behavioural forma mentis networks (BFMNs) to investigate the STEM-focused mindset, i.e. ways of associating and perceiving ideas, of 177 human participants and 177 artificial humans simulated by GPT-3.5. Participants were split in 3 groups - trainees, experts and academics - to compare the influence of expertise level on their mindset. The results revealed that human forma mentis networks exhibited significantly higher clustering coefficients compared to GPT-3.5, indicating that human mindsets displayed a tendency to form and close triads of conceptual associations while recollecting STEM ideas. Human experts, in particular, demonstrated robust clustering coefficients, reflecting better integration of STEM concepts into their cognitive networks. In contrast, GPT-3.5 produced sparser mindsets. Furthermore, both human and GPT mindsets framed mathematics in neutral or positive terms, differently from STEM high schoolers, researchers and other large language models sampled in other works. This research contributes to understanding how mindset structure can provide cognitive insights about memory structure and machine limitations.

cs.CL↗

arXiv2024

Large language models surpass human experts in predicting neuroscience results

Scientific discoveries often hinge on synthesizing decades of research, a task that potentially outstrips human information processing capacities. Large language models (LLMs) offer a solution. LLMs trained on the vast scientific literature could potentially integrate noisy yet interrelated findings to forecast novel results better than human experts. To evaluate this possibility, we created BrainBench, a forward-looking benchmark for predicting neuroscience results. We find that LLMs surpass experts in predicting experimental outcomes. BrainGPT, an LLM we tuned on the neuroscience literature, performed better yet. Like human experts, when LLMs were confident in their predictions, they were more likely to be correct, which presages a future where humans and LLMs team together to make discoveries. Our approach is not neuroscience-specific and is transferable to other knowledge-intensive endeavors.

q-bio.NC↗

arXiv2024

Network Representation of Higher-Order Interactions Based on Information Dynamics

Many complex systems in science and engineering are modeled as networks whose nodes and links depict the temporal evolution of each system unit and the dynamic interaction between pairs of units, which are assessed respectively using measures of auto- and cross-correlation or variants thereof. However, a growing body of work is documenting that this standard network representation can neglect potentially crucial information shared by three or more dynamic processes in the form of higher-order interactions (HOIs). While several measures, mostly derived from information theory, are available to assess HOIs in network systems mapped by multivariate time series, none of them is able to provide a compact and detailed representation of higher-order interdependencies. In this work, we fill this gap by introducing a framework for the assessment of HOIs in dynamic network systems at different levels of resolution. The framework is grounded on the dynamic implementation of the O-information, a new measure assessing HOIs in dynamic networks, which is here used together with its local counterpart and its gradient to quantify HOIs respectively for the network as a whole, for each link, and for each node. The integration of these measures into the conventional network representation results in a tool for the representation of HOIs as networks, which is defined formally using measures of information dynamics, implemented in its linear version by using vector regression models and statistical validation techniques, illustrated in simulated network systems, and finally applied to an illustrative example in the field of network physiology.

stat.ME↗

arXiv2024

Information-Theoretic Progress Measures reveal Grokking is an Emergent Phase Transition

This paper studies emergent phenomena in neural networks by focusing on grokking where models suddenly generalize after delayed memorization. To understand this phase transition, we utilize higher-order mutual information to analyze the collective behavior (synergy) and shared properties (redundancy) between neurons during training. We identify distinct phases before grokking allowing us to anticipate when it occurs. We attribute grokking to an emergent phase transition caused by the synergistic interactions between neurons as a whole. We show that weight decay and weight initialization can enhance the emergent phase.

cs.LG↗

arXiv2024

Disentangling high order effects in the transfer entropy

Transfer Entropy (TE), the primary method for determining directed information flow within a network system, can exhibit bias - either in deficiency or excess - during both pairwise and conditioned calculations, owing to high-order dependencies among the dynamic processes under consideration and the remaining processes in the system used for conditioning. Here, we propose a novel approach. Instead of conditioning TE on all network processes except the driver and target, as in its fully conditioned version, or not conditioning at all, as in the pairwise approach, our method searches for both the multiplets of variables that maximize information flow and those that minimize it. This provides a decomposition of TE into unique, redundant, and synergistic atoms. Our approach enables the quantification of the relative importance of high-order effects compared to pure two-body effects in information transfer between two processes, while also highlighting the processes that contribute to building these high-order effects alongside the driver. We demonstrate the application of our approach in climatology by analyzing data from El Niño and the Southern Oscillation.

physics.data-an↗

arXiv2024

Assessing High-Order Links in Cardiovascular and Respiratory Networks via Static and Dynamic Information Measures

The network representation is becoming increasingly popular for the description of cardiovascular interactions based on the analysis of multiple simultaneously collected variables. However, the traditional methods to assess network links based on pairwise interaction measures cannot reveal high-order effects involving more than two nodes, and are not appropriate to infer the underlying network topology. To address these limitations, here we introduce a framework which combines the assessment of high-order interactions with statistical inference for the characterization of the functional links sustaining physiological networks. The framework develops information-theoretic measures quantifying how two nodes interact in a redundant or synergistic way with the rest of the network, and employs these measures for reconstructing the functional structure of the network. The measures are implemented for both static and dynamic networks mapped respectively by random variables and random processes using plug-in and model-based entropy estimators. The validation on theoretical and numerical simulated networks documents the ability of the framework to represent high-order interactions as networks and to detect statistical structures associated to cascade, common drive and common target effects. The application to cardiovascular networks mapped by the beat-to-beat variability of heart rate, respiration, arterial pressure, cardiac output and vascular resistance allowed noninvasive characterization of several mechanisms of cardiovascular control operating in resting state and during orthostatic stress. Our approach brings to new comprehensive assessment of physiological interactions and complements existing strategies for the classification of pathophysiological states.

stat.ME↗

arXiv2023

brainlife.io: A decentralized and open source cloud platform to support neuroscience research

Neuroscience research has expanded dramatically over the past 30 years by advancing standardization and tool development to support rigor and transparency. Consequently, the complexity of the data pipeline has also increased, hindering access to FAIR (Findable, Accessible, Interoperabile, and Reusable) data analysis to portions of the worldwide research community. brainlife.io was developed to reduce these burdens and democratize modern neuroscience research across institutions and career levels. Using community software and hardware infrastructure, the platform provides open-source data standardization, management, visualization, and processing and simplifies the data pipeline. brainlife.io automatically tracks the provenance history of thousands of data objects, supporting simplicity, efficiency, and transparency in neuroscience research. Here brainlife.io's technology and data services are described and evaluated for validity, reliability, reproducibility, replicability, and scientific utility. Using data from 4 modalities and 3,200 participants, we demonstrate that brainlife.io's services produce outputs that adhere to best practices in modern neuroscience research.

cs.DC↗

arXiv2023

Frontal effective connectivity increases with task demands and time on task: a Dynamic Causal Model of electrocorticogram in macaque monkeys

We apply Dynamic Causal Models to electrocorticogram recordings from two macaque monkeys performing a problem-solving task that engages working memory, and induces time-on-task effects. We thus provide a computational account of changes in effective connectivity within two regions of the fronto-parietal network, the dorsolateral prefrontal cortex and the pre-supplementary motor area. We find that forward connections between the two regions increased in strength when task demands increased, and as the experimental session progressed. Similarities in the effects of task demands and time on task allow us to interpret changes in frontal connectivity in terms of increased attentional effort allocation that compensates cognitive fatigue.

q-bio.NC↗

arXiv2022

Higher-order mutual information reveals synergistic sub-networks for multi-neuron importance

Quantifying which neurons are important with respect to the classification decision of a trained neural network is essential for understanding their inner workings. Previous work primarily attributed importance to individual neurons. In this work, we study which groups of neurons contain synergistic or redundant information using a multivariate mutual information method called the O-information. We observe the first layer is dominated by redundancy suggesting general shared features (i.e. detecting edges) while the last layer is dominated by synergy indicating local class-specific features (i.e. concepts). Finally, we show the O-information can be used for multi-neuron importance. This can be demonstrated by re-training a synergistic sub-network, which results in a minimal change in performance. These results suggest our method can be used for pruning and unsupervised representation learning.

cs.LG↗