Integrating Spiking Neural Networks in Neurobiology and Computer Science

Integrative Approaches in Spiking Neural Networks: Bridging Machine Learning, Computer Science, and Neurobiology (11:15-11:45) 

Matej Mertik¹ and Maciej Wielgosz^2,3

¹Alma Mater Europaea ECM, Maribor, Slovenia

²Faculty of Computer Science, Electronics and Telecommunications, Department of Electronics, AGH University of Science and Technology, Kraków, Poland

³ACC Cyfronet AGH, Kraków, Poland

Abstract: 

Spiking Neural Networks (SNNs) stand at the intersection of machine learning (ML), computer science (CS), and neurobiology, promising to revolutionize computational paradigms by mimicking the temporal dynamics of biological neural systems. This presentation explores the landscape of SNNs, emphasizing the synthesis of concepts from ML, such as learning algorithms and frameworks like BindsNET and SpikeJelly, with the structural and functional insights from neurobiology. We discuss the role of computer science in developing simulation tools like NEST, which enable the mapping of neural architectures. The contribution of neurobiology is underscored by providing biological fidelity to models, ifluencing both architecture and function. Moreover, we delve into the emerging field of neuromorphic engineering (EE), which aims to translate the computational efficiency of SNNs into hardware implementations. By converging these fields, SNNs hold the potential for creating more efficient, adaptive, and biologically realistic computing systems. The presentation concludes with a discussion on the current challenges and future directions in SNN research, outlining a collaborative path forward for these intertwined disciplines.

​​Exploring the Versatility of Multielectrode Array Systems in Ex Vivo Neural Recordings​  (11:45-12:00) 

K. Przybylska, A. Trenk and A. Blasiak 

Department of Neurophysiology and Chronobiology, Institute of Zoology and Biomedical Research, Jagiellonian University, Cracow, Poland

Abstract: Multielectrode Array (MEA) systems present one of the several methods of data acquisition from the brain, offering a window into neural dynamics. This presentation will provide a comprehensive overview of MEA systems, positioning them as a robust tool in neuroscience research. Focusing on ex vivo recordings, the discussion will highlight the diverse applications of MEA technology, particularly in conjunction with other advanced techniques, such as optogenetics and chemogenetics. Furthermore, the intricacies of data analysis and interpretation will be delved into, providing practical insights for researchers.  
 
By elucidating the functional aspects of MEA technology, this talk aims to enhance understanding of this technique and emphasize its potential as a valuable resource for researchers seeking comprehensive neural activity data

Funding: ​​This research was supported by a research grant from the National Science Centre, Poland (UMO-2023/49/B/NZ4/01885).​ 

​​Enriching Spiking Neural Network models with properties of neurotransmitters​ (12:00-12:15)

​​Szymon Mazurek^1,2

​​¹Brain and More Lab, Sano Centre for Computational Medicine, Cracow, Poland 
²Department of Electronics and Telecommunications, AGH University of Krakow, Cracow, Poland​ 

Abstract: Modern artificial neural networks recently reached stellar performance across various tasks, previously deemed as unsolvable by machine models. Their principle of operation, however, resembles the biological intelligence phenomena poorly. Spiking Neural Networks show promise in bridging this gap. Yet sAll, they need to be completed in terms of biological plausibility, operating on simplified neuron models and mostly Hebbian-based learning rules. Recent advances in neuroscience shed more light on the principles of brain operation, showing new ways of modeling intelligent systems in sillico. Today I would like to show how neuromodulatory mechanisms can be included in these networks, possibly improving their performance in various environments.​   

Funding: The publication was created within the project of the Minister of Science and Higher Education “Support for the activity of Centers of Excellemce established in Poland under Horizon 2020” on the basis of the contract number MEiN/2023?DIR/3796. This publication is supported by the European Union’s Horizon 2020 researcg and innovation programme under grant agreement Sano No 857533. This publication is supported by Sano project carried out within the International Research Agendas programme of the Foundation for Polish Science, co-financed by the European Union under the European Regional Development Fund.

​​Training Biologically Plausible Neural Networks: Current Challenges​ (12:15-12:30) 

​​Joan Falcó-Roget​ 

​​Computer Vision Team, Brain & More Lab, Sano Centre for Computational Medicine, Kraków, Poland​ 

Abstract: The successful deep artificial neural networks barely resemble how networks in the mammal brain operate. Behavioral experiments and computational modeling of those offer the possibility to hinder some of these differences. First, I will discuss the basic properties of cortical networks and how we can incorporate them into biologically plausible mathematical models. Second, I will emphasize how the training of these network models remains an open challenge. Moreover, these challenges increase enormously if we incorporate further biological constraints into the learning process; for example, Hebbian-based and/or dopamine-based learning rules. Finally, and most importantly, I will discuss how we could bypass these constraints to obtain network models to mimic information processing in real brains and how they can be used to understand how behavior and cognition emerge from low-level computations.

Funding:  The publication was created within the project of the Minister of Science and Higher Education "Support for the activity of Centers of Excellence established in Poland under Horizon 2020" on the basis of the contract number MEiN/2023/DIR/3796. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 857533. This publication is supported by Sano project carried out within the International Research Agendas programme of the Foundation for Polish Science, co-financed by the European Union under the European Regional Development Fund.​

​​Modeling brain activity with Spiking Neural Networks​ (12:30-12:45) 

​​Jan K. Argasiński​ 

​​Jagiellonian University in Krakow & Sano - Centre for Computational Medicine, Krakow​, Poland 

Abstract: Spiking Neural Networks (SNNs) represent a sophisticated approach to modelling brain activity, closely mimicking the dynamics of biological neural networks. By incorporating the fundamental principles of how neurons spike and communicate, SNNs provide a more realistic representation of neural processing compared to traditional artificial neural networks. When it comes to integrating multielectrode array recordings from rat brain areas into computational modeling, SNNs play a crucial role. These recordings offer rich, detailed data about neuronal activity patterns, including the timing of spikes, which is critical for SNNs. By inputting this data into SNNs, researchers can simulate how specific brain areas process information. This approach allows for a deeper understanding of complex neural mechanisms and can lead to advancements in neuroscientific research, offering insights into brain function, neural disorders, and potential therapeutic strategies. The alignment of SNNs with real neural data enhances the accuracy and relevance of computational models in neuroscience.​