Staff member


Xerxes Arsiwalla

Postdoctoral Researcher
Synthetic, Perceptive, Emotive and Cognitive Systems (SPECS)
xarsiwalla@ibecbarcelona.eu

Staff member publications

Freire, I. T., Moulin-Frier, C., Sanchez-Fibla, M., Arsiwalla, X. D., Verschure, P., (2018). Modeling the formation of social conventions in multi-agent populations ARXIV Computer Science, (Multiagent Systems), 1-30

In order to understand the formation of social conventions we need to know the specific role of control and learning in multi-agent systems. To advance in this direction, we propose, within the framework of the Distributed Adaptive Control (DAC) theory, a novel Control-based Reinforcement Learning architecture (CRL) that can account for the acquisition of social conventions in multi-agent populations that are solving a benchmark social decision-making problem. Our new CRL architecture, as a concrete realization of DAC multi-agent theory, implements a low-level sensorimotor control loop handling the agent's reactive behaviors (pre-wired reflexes), along with a layer based on model-free reinforcement learning that maximizes long-term reward. We apply CRL in a multi-agent game-theoretic task in which coordination must be achieved in order to find an optimal solution. We show that our CRL architecture is able to both find optimal solutions in discrete and continuous time and reproduce human experimental data on standard game-theoretic metrics such as efficiency in acquiring rewards, fairness in reward distribution and stability of convention formation.

Keywords: Computer Science, Multiagent Systems


Zamora, R., Korff, S., Mi, Q., Barclay, D., Schimunek, L., Zucca, R., Arsiwalla, X. D., Simmons, R. L., Verschure, P., Billiar, T. R., Vodovotz, Y., (2018). A computational analysis of dynamic, multi-organ inflammatory crosstalk induced by endotoxin in mice PLoS Computational Biology 14, (11), e1006582

Bacterial lipopolysaccharide (LPS) induces an acute inflammatory response across multiple organs, primarily via Toll-like receptor 4 (TLR4). We sought to define novel aspects of the complex spatiotemporal dynamics of LPS-induced inflammation using computational modeling, with a special focus on the timing of pathological systemic spillover. An analysis of principal drivers of LPS-induced inflammation in the heart, gut, lung, liver, spleen, and kidney to assess organ-specific dynamics, as well as in the plasma (as an assessment of systemic spillover), was carried out using data on 20 protein-level inflammatory mediators measured over 0-48h in both C57BL/6 and TLR4-null mice. Using a suite of computational techniques, including a time-interval variant of Principal Component Analysis, we confirm key roles for cytokines such as tumor necrosis factor-α and interleukin-17A, define a temporal hierarchy of organ-localized inflammation, and infer the point at which organ-localized inflammation spills over systemically. Thus, by employing a systems biology approach, we obtain a novel perspective on the time- and organ-specific components in the propagation of acute systemic inflammation.


Arsiwalla, Xerxes D., Verschure, Paul, (2018). Measuring the complexity of consciousness Frontiers in Neuroscience 12, (424), Article 424

The grand quest for a scientific understanding of consciousness has given rise to many new theoretical and empirical paradigms for investigating the phenomenology of consciousness as well as clinical disorders associated to it. A major challenge in this field is to formalize computational measures that can reliably quantify global brain states from data. In particular, information-theoretic complexity measures such as integrated information have been proposed as measures of conscious awareness. This suggests a new framework to quantitatively classify states of consciousness. However, it has proven increasingly difficult to apply these complexity measures to realistic brain networks. In part, this is due to high computational costs incurred when implementing these measures on realistically large network dimensions. Nonetheless, complexity measures for quantifying states of consciousness are important for assisting clinical diagnosis and therapy. This article is meant to serve as a lookup table of measures of consciousness, with particular emphasis on clinical applicability. We consider both, principle-based complexity measures as well as empirical measures tested on patients. We address challenges facing these measures with regard to realistic brain networks, and where necessary, suggest possible resolutions. We address challenges facing these measures with regard to realistic brain networks, and where necessary, suggest possible resolutions.

Keywords: Consciousness in the Clinic, Computational neuroscience, Complexity measures, Clinical Neuroscience, Measures of consciousness


Arsiwalla, Xerxes, Signorelli, Camilo M., Puigbo, Jordi-Ysard, Freire, Ismael, Verschure, P., (2018). Are brains computers, emulators or simulators? Biomimetic and Biohybrid Systems 7th International Conference, Living Machines 2018 (Lecture Notes in Computer Science) , Springer International Publishing (Paris, France) 10928, 11-15

There has been intense debate on the question of whether the brain is a computer. If so, that challenge is to show that all cognitive processes can be described by algorithms running on a universal Turing machine. By extension that implies consciousness is a computational process. Both Penrose and Searle have vehemently argued against this view, proposing that consciousness is a fundamentally non-computational process. Even proponents of the brain as a computer metaphor such a Dennett agree that the organizational architecture of the brain is unlike any computing system ever conceived, possibly alluding to non-classical computational processes. The latter class of processes veer away from any program that can be encoded by Church’s lambda calculus. In fact, such a program would have to be based on non-classical logic (either semi-classical or quantum). But quantum logic or machines that might implement them typically are not meant for solving the same type of problems that a classical computer solves (nor are they necessarily faster for any given problem). We will argue that machines implementing non-classical logic might be better suited for simulation rather than computation (a la Turing). It is thus reasonable to pit simulation as an alternative to computation and ask whether the brain, rather than computing, is simulating a model of the world in order to make predictions and guide behavior. If so, this suggests a hardware supporting dynamics more akin to a quantum many-body field theory.


Moulin-Frier, C., Puigbò, J. Y., Arsiwalla, X. D., Sanchez-Fibla, M., Verschure, P., (2018). Embodied artificial intelligence through distributed adaptive control: An integrated framework ICDL-EpiRob 2017 7th Joint IEEE International Conference on Development and Learning and on Epigenetic Robotics , IEEE (Lisbon, Portugal) , 324-330

In this paper, we argue that the future of Artificial Intelligence research resides in two keywords: integration and embodiment. We support this claim by analyzing the recent advances in the field. Regarding integration, we note that the most impactful recent contributions have been made possible through the integration of recent Machine Learning methods (based in particular on Deep Learning and Recurrent Neural Networks) with more traditional ones (e.g. Monte-Carlo tree search, goal babbling exploration or addressable memory systems). Regarding embodiment, we note that the traditional benchmark tasks (e.g. visual classification or board games) are becoming obsolete as state-of-the-art learning algorithms approach or even surpass human performance in most of them, having recently encouraged the development of first-person 3D game platforms embedding realistic physics. Building on this analysis, we first propose an embodied cognitive architecture integrating heterogeneous subfields of Artificial Intelligence into a unified framework. We demonstrate the utility of our approach by showing how major contributions of the field can be expressed within the proposed framework. We then claim that benchmarking environments need to reproduce ecologically-valid conditions for bootstrapping the acquisition of increasingly complex cognitive skills through the concept of a cognitive arms race between embodied agents.

Keywords: Cognitive Architectures, Embodied Artificial Intelligence, Evolutionary Arms Race, Unified Theories of Cognition


Freire, Ismael, Puigbo, J., Arsiwalla, Xerxes, Verschure, Paul, (2018). Modeling the opponent’s action using control-based reinforcement learning Biomimetic and Biohybrid Systems 7th International Conference, Living Machines 2018 (Lecture Notes in Computer Science) , Springer International Publishing (Paris, France) 10928, 179-186

In this paper, we propose an alternative to model-free reinforcement learning approaches that recently have demonstrated Theory-of-Mind like behaviors. We propose a game theoretic approach to the problem in which pure RL has demonstrated to perform below the standards of human-human interaction. In this context, we propose alternative learning architectures that complement basic RL models with the ability to predict the other’s actions. This architecture is tested in different scenarios where agents equipped with similar or varying capabilities compete in a social game. Our different interaction scenarios suggest that our model-based approaches are especially effective when competing against models of equivalent complexity, in contrast to our previous results with more basic predictive architectures. We conclude that the evolution of mechanisms that allow for the control of other agents provide different kinds of advantages that can become significant when interacting with different kinds of agents. We argue that no single proposed addition to the learning architecture is sufficient to optimize performance in these scenarios, but a combination of the different mechanisms suggested is required to achieve near-optimal performance in any case.


Arsiwalla, X. D., Pacheco, D., Principe, A., Rocamora, R., Verschure, P., (2018). A temporal estimate of integrated information for intracranial functional connectivity Artificial Neural Networks and Machine Learning (Lecture Notes in Computer Science) 27th International Conference on Artificial Neural Networks (ICANN 2018) , Springer, Cham (Rhodes, Greece) 11140, 403-412

A major challenge in computational and systems neuroscience concerns the quantification of information processing at various scales of the brain’s anatomy. In particular, using human intracranial recordings, the question we ask in this paper is: How can we estimate the informational complexity of the brain given the complex temporal nature of its dynamics? To address this we work with a recent formulation of network integrated information that is based on the Kullback-Leibler divergence between the multivariate distribution on the set of network states versus the corresponding factorized distribution over its parts. In this work, we extend this formulation for temporal networks and then apply it to human brain data obtained from intracranial recordings in epilepsy patients. Our findings show that compared to random re-wirings of the data, functional connectivity networks, constructed from human brain data, score consistently higher in the above measure of integrated information. This work suggests that temporal integrated information may indeed be a good starting point as a future measure of cognitive complexity.

Keywords: Brain networks, Complexity measures, Computational neuroscience, Functional connectivity


Puigbò, J. Y., Arsiwalla, X. D., Verschure, P., (2018). Challenges of machine learning for living machines Biomimetic and Biohybrid Systems 7th International Conference, Living Machines 2018 (Lecture Notes in Computer Science) , Springer International Publishing (Paris, France) 10928, 382-386

Machine Learning algorithms (and in particular Reinforcement Learning (RL)) have proved very successful in recent years. These have managed to achieve super-human performance in many different tasks, from video-games to board-games and complex cognitive tasks such as path-planning or Theory of Mind (ToM) on artificial agents. Nonetheless, this super-human performance is also super-artificial. Despite some metrics are better than what a human can achieve (i.e. cumulative reward), in less common metrics (i.e. time to learning asymptote) the performance is significantly worse. Moreover, the means by which those are achieved fail to extend our understanding of the human or mammal brain. Moreover, most approaches used are based on black-box optimization, making any comparison beyond performance (e.g. at the architectural level) difficult. In this position paper, we review the origins of reinforcement learning and propose its extension with models of learning derived from fear and avoidance behaviors. We argue that avoidance-based mechanisms are required when training on embodied, situated systems to ensure fast and safe convergence and potentially overcome some of the current limitations of the RL paradigm.

Keywords: Avoidance, Neural networks, Reinforcement learning


Freire, I. T., Arsiwalla, X. D., Puigbò, J. Y., Verschure, P., (2018). Limits of multi-agent predictive models in the formation of social conventions Frontiers in Artificial Intelligence and Applications (ed. Falomir, Z., Gibert, K., Plaza, E.), IOS Press (Amsterdam, The Netherlands) Volume 308: Artificial Intelligence Research and Development, 297-301

A major challenge in cognitive science and AI is to understand how intelligent agents might be able to predict mental states of other agents during complex social interactions. What are the computational principles of such a Theory of Mind (ToM)? In previous work, we have investigated hypotheses of how the human brain might realize a ToM of other agents in a multi-agent social scenario. In particular, we have proposed control-based cognitive architectures to predict the model of other agents in a game-theoretic task (Battle of the Exes). Our multi-layer architecture implements top-down predictions from adaptive to reactive layers of control and bottom-up error feedback from reactive to adaptive layers. We tested cooperative and competitive strategies among different multi-agent models, demonstrating that while pure RL leads to reasonable efficiency and fairness in social interactions, there are other architectures that can perform better in specific circumstances. However, we found that even the best predictive models fall short of human data in terms of stability of social convention formation. In order to explain this gap between humans and predictive AI agents, in this work we propose introducing the notion of trust in the form of mutual agreements between agents that might enhance stability in the formation of conventions such as turn-taking.

Keywords: Cognitive Architectures, Game Theory, Multi-Agent Models, Reinforcement Learning, Theory of Mind


Arsiwalla, X. D., Signorelli, C. M., Puigbo, J. Y., Freire, I. T., Verschure, P., (2018). What is the physics of intelligence? Frontiers in Artificial Intelligence and Applications (ed. Falomir, Z., Gibert, K., Plaza, E.), IOS Press (Amsterdam, The Netherlands) Volume 308: Artificial Intelligence Research and Development, 283-286

In the absence of a first-principles definition, the concept of intelligence is often specified in terms of its phenomenological functions as a capacity or ability to solve problems autonomously. Whenever an agent, biological or artificial, possesses this ability, it is considered intelligent, otherwise not. While this description serves as a useful correlate of intelligence, it is far from a principled explanation that provides a general, yet precise definition along with predictions of mechanisms leading to intelligent behavior. We do not want an explanation to depend on any functionality that itself might be a consequence of intelligence. A possible conceptualization of a function-free approach might be to formulate the concept in terms of dynamical information complexity. This constitute a first step towards a statistical mechanics theory of intelligence. In this paper, we outline the steps towards a physics-based definition of intelligence.

Keywords: Complexity, Information Theory, Physics of Intelligence


Hindriks, Rikkert, Schmiedt, Joscha, Arsiwalla, Xerxes D., Peter, Alina, Verschure, Paul F. M. J., Fries, Pascal, Schmid, Michael C., Deco, Gustavo, (2017). Linear distributed source modeling of local field potentials recorded with intra-cortical electrode arrays PLoS ONE 12, (12), e0187490

Planar intra-cortical electrode (Utah) arrays provide a unique window into the spatial organization of cortical activity. Reconstruction of the current source density (CSD) underlying such recordings, however, requires “inverting” Poisson’s equation. For inter-laminar recordings, this is commonly done by the CSD method, which consists in taking the second-order spatial derivative of the recorded local field potentials (LFPs). Although the CSD method has been tremendously successful in mapping the current generators underlying inter-laminar LFPs, its application to planar recordings is more challenging. While for inter-laminar recordings the CSD method seems reasonably robust against violations of its assumptions, is it unclear as to what extent this holds for planar recordings. One of the objectives of this study is to characterize the conditions under which the CSD method can be successfully applied to Utah array data. Using forward modeling, we find that for spatially coherent CSDs, the CSD method yields inaccurate reconstructions due to volume-conducted contamination from currents in deeper cortical layers. An alternative approach is to “invert” a constructed forward model. The advantage of this approach is that any a priori knowledge about the geometrical and electrical properties of the tissue can be taken into account. Although several inverse methods have been proposed for LFP data, the applicability of existing electroencephalographic (EEG) and magnetoencephalographic (MEG) inverse methods to LFP data is largely unexplored. Another objective of our study therefore, is to assess the applicability of the most commonly used EEG/MEG inverse methods to Utah array data. Our main conclusion is that these inverse methods provide more accurate CSD reconstructions than the CSD method. We illustrate the inverse methods using event-related potentials recorded from primary visual cortex of a macaque monkey during a motion discrimination task.


Moulin-Frier, C., Puigbò, J.-Y., Arsiwalla, Xerxes D., Martì Sanchez-Fibla, M., Verschure, Paul F. M. J., (2017). Embodied artificial intelligence through distributed adaptive control: An integrated framework 7th Joint IEEE International Conference on Development and Learning and on Epigenetic Robotics (ICDL-Epirob 2017) , IEEE (Lisbon, Portugal) , 1-8

In this paper, we argue that the future of Artificial Intelligence research resides in two keywords: integration and embodiment. We support this claim by analyzing the recent advances of the field. Regarding integration, we note that the most impactful recent contributions have been made possible through the integration of recent Machine Learning methods (based in particular on Deep Learning and Recurrent Neural Networks) with more traditional ones (e.g. Monte-Carlo tree search, goal babbling exploration or addressable memory systems). Regarding embodiment, we note that the traditional benchmark tasks (e.g. visual classification or board games) are becoming obsolete as state-of-the-art learning algorithms approach or even surpass human performance in most of them, having recently encouraged the development of first-person 3D game platforms embedding realistic physics. Building upon this analysis, we first propose an embodied cognitive architecture integrating heterogenous sub-fields of Artificial Intelligence into a unified framework. We demonstrate the utility of our approach by showing how major contributions of the field can be expressed within the proposed framework. We then claim that benchmarking environments need to reproduce ecologically-valid conditions for bootstrapping the acquisition of increasingly complex cognitive skills through the concept of a cognitive arms race between embodied agents.

Keywords: Cognitive Architectures, Embodied Artificial Intelligence, Evolutionary Arms Race, Unified Theories of Cognition