NeuroMat’s strategic directions for the next two years

Jan 29, 2016

NeuroMat has launched a strategic roadmap to direct research, technology-transfer and scientific-dissemination activities until 2018. NeuroMat is the São Paulo Research Foundation (FAPESP)’s Research, Innovation and Dissemination Center for Neuromathematics, created in 2013 and hosted at the University of São Paulo, Brazil. The general goal of this project that probabilist Antonio Galves leads since its inception is to devise the new mathematics that are needed to make sense of the large, continuous inflow of research results neurobiology generates.

The strategic roadmap encompasses medium-term goals in the three areas of the RIDC. “The fulfillment of these goals is imperative. NeuroMat was granted the opportunity to tackle a research agenda of extreme importance that misleadingly appeared intractable to many scholars and is mandated to speed up progress in the mathematical understanding of the brain,” said Galves. NeuroMat principal investigators took part of the setting of the medium-term goals, under the coordination of the RIDC executive manager, Magda Chang. The setting of these medium-term goals resulted from a critical evaluation of NeuroMat’s activities assessed in NeuroMat's reports of activities and inputs from FAPESP's evaluation committee that visited the research center at the end of last November, enlightened by a reassessment of NeuroMat’s initial research project.

Main research goals for the next two years include improving NeuroMat's new class of stochastic processes to model systems of interacting neurons and developing NeuroMat’s new class of stochastic processes driven by context-tree models. The modeling of systems of interacting neurons is at the core of the RIDC project and was introduced in the article written by Galves and Löcherbach “Infinite systems of interacting chains with memory of variable length - a stochastic model for biological neural nets,” in 2013. Main achievements in this research direction were addressing the issue of hydrodynamic limits for this class of stochastic processes, the study of the long term behavior of the stochastic processes in this class of models and the presentation of new perfect simulation methods for this class of stochastic processes.

Research perspectives for the next two years to cope with the challenge of modeling systems of interacting neurons involve rigorously defining brain plasticity and incorporating it in the model, understanding the model in the finite case, making computational simulations of the model —with the caveat that it rests upon a nondeterministic account of neural activity— and understanding the properties of the random graphs that should be used in the modeling of interacting neurons. The mathematical advances in the development of the Galves-Löcherbach model will necessarily be accompanied by the development of the statistical tools required by this new class of stochastic processes, which is an ongoing effort within NeuroMat.

Computational simulations of the Galves-Löcherbach model is the core goal of NeuroMat’s High-Performance Computational Center, that should be launched in April at the University of São Paulo, Ribeirão Preto campus. Simulating neural systems should be a major discussion point in a forthcoming conference NeuroMat’s research team will organize at the end of April, around “Frontiers of Brain Theory” (more information to be released soon at the Events tab on NeuroMat’s website). The properties of the random graphs that should be used in the modeling of interacting neurons have been dealt with at the workshop “Random Graphs in the Brain,” in November, 2015, and should be topic of a future workshop provisorily called “Random Structures in the Brain,” with details to be announced.

The research agenda around retrieving a context tree from EEG data remains a strategic goal of the RIDC. This agenda relates to the NeuroMat conjecture that the brain depends on statistical model selection to make sense of stimuli, incorporating context-tree models to bring experimental evidences of this conjecture. A paper on the mathematical class of models that was used to retrieve a context tree from EEG data has been submitted, and a paper on the neurobiological approach on the experiment that brought evidence that supported the conjecture. The idea that the brain does statistical model selection has led to a new experimental protocol, that combines context-tree model and learning processes. This new experimental protocol was recently extended to a protocol based on behavorial data retrieved from a computer game that was released to the public and was scientifically validated in 2015, and could be the basis of research on rehabilitation of patients with neurological diseases. Advances in rehabilitation could also lead to a new line of activity within the technology-transfer team.

The linking of high-level scientific endeavors, particularly what is accomplished within NeuroMat, and broader audiences remains the main mission of the RIDC's dissemination team. This mission has geared dissemination activities within the research center but also a scientific discussion on the best strategies to perform such linking, because the development of new scientific areas, such as Neuromathematics, imposes a set of challenges to communicating and disseminating news on such development. "Results are in the making, conceptual frameworks remain under construction and often at a high level of abstraction, levels of uncertainty are generally unavoidable. This scenario leads to a dilemma to any serious attempt of communicating and disseminating cutting-edge and ongoing scientific work: to rely on well-established repertoires of communication with the risk of systematically failing to keep track of standards these repertoires require, or to advance new means of communicating and disseminating in parallel to the development of new scientific areas. This second part of the dilemma –which suggests including the science of communication in the realm of broader scientific agendas– involves high risk-taking in the process of reporting scientific enterprises," as a report on the NeuroMat event "New Forms of Scientific Journalism and the RIDC NeuroMat's Work" put it.

A massive open online course on the conjecture that the brain does statistical model selection remains the leading effort of NeuroMat’s dissemination team for the next two years. This effort has brought together members of the research and dissemination teams, as long as independent researchers, to create a multimedia educational project. This course will be initially offered to public-school teachers, eventually contributing to improving the content on data analysis, mathematics and neurobiology, that are covered in the curriculum.

The communication work of NeuroMat’s dissemination team involves increasing the RIDC media exposure and improving the quality of the content that is released from the center. This content includes NeuroMat’s monthly newsletter, web portal and Facebook page. The monthly newsletter has been the major source of information from the RIDC for two years, providing extensive coverage of main achievements and directions of the center. The NeuroMat web portal was launched in early February 2014, and brings up all relevant updates in the research, technology transfer and dissemination activities. Publications are in English and Portuguese. Around 8,500 different users –51% of whom were identified as non-Brazilian users– have visited NeuroMat’s webpage since its inception, with 61,000 page views. The NeuroMat Facebook page brings up short texts and media content on a daily basis. Since its creation (9/20/2014), the page has reached approximately 560 likes. NeuroMat’s dissemination team expects to start the release of periodic podcasts on topics pertaining to Neuromathematics, that could eventually be disseminated on radio broadcasts.

The NeuroMat dissemination team intends to intensify its Wikipedia initiative. Responding to the call of Wikipedia’s Year of Science, the expectation is that content on Neuromathematics on the Portuguese and English versions of Wikipedia is improved.

Three movies are currently being produced within NeuroMat —on the experimental protocol that brought evidence on the conjecture that the brain does statistical model selection, on open science and on the Latin American School on Computational Neuroscience. The expectation is that at least two of these videos will be released in 2016.

The development of open-source computational tools that the scientific questions to advance Neuromathematics require is the core mission of NeuroMat's technology-transfer team. Such development involves both operational and scientific challenges, mostly pertaining to the science of database structuring and modeling.

The NeuroMat technology-transfer team intends to level up the ongoing effort to create a database for neurobiological data. The first step of this effort has been geared towards the development of the Neuroscience Experiments System (NES), an open-source tool to manage clinical data gathered in hospitals and research institutions. The development of NES has been directed by NeuroMat’s stand on promoting open data, that has been considered a key component in advancing clinical treatments and has led to innovative organizational designs among software-development team members. After the first stage of developing NES, the technology-transfer team expects to document the protocols for experimental data collection, so that it becomes an input in the database, and to share electronically the database. The clinical and research team at the Institute of Neurology Deolindo Couto, from the Federal University of Rio de Janeiro (INDC/UFRJ), pioneered the use of NES and NeuroMat’s database, and has had an important role in the defining features of the software.

NeuroMat’s technology-transfer strategy is expected to join a specialized international network on the development of tools for neurobiology, that is organized by the International Neuroinformatics Coordinating Facility (INCF). The expectation is that this could intensify exchanges and improve the programming speed within NeuroMat and centers that focus on theoretical neuroscience.

(This document was produced with inputs from NeuroMat members and as other inputs are received some updates might be made.)

This piece is part of NeuroMat's Newsletter #24. Read more here

Share on Twitter Share on Facebook