A weighted generative model of the human connectome

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Июнь 25, 2023

Abstract Probabilistic generative network models have offered an exciting window into the constraints governing human connectome’s organization. In particular, they highlighted economic context of formation and special roles that physical geometry self-similarity likely play in determining topology. However, a critical limitation these is do not consider strength anatomical connectivity between regions. This significantly limits their scope to answer neurobiological questions. The current work draws inspiration from principle redundancy reduction develop novel weighted model. model significant advance because it only incorporates theoretical advancements previous models, but also has ability capture dynamic strengthening or weakening connections over time. Using state-of-the-art Convex Optimization Modelling for Microstructure-Informed Tractography (COMMIT) approach, sample children adolescents ( n = 88, aged 8 18 years), we show this can accurately approximate simultaneously topology edge-weights connectome (specifically, MRI signal fraction attributed axonal projections). We achieve at both sparse dense densities. Generative fits are comparable to, many cases better than, published findings simulating absence weights. Our implications future research by providing new avenues exploring normative developmental trends, neural computation wider conceptual economics connectomics supporting functioning.

Язык: Английский

---

Cognition Without Neural Representation: Dynamics of a Complex System

Frontiers in Psychology, Год журнала: 2022, Номер 12

Опубликована: Янв. 12, 2022

This paper proposes an account of neurocognitive activity without leveraging the notion neural representation. Neural representation is a concept that results from assuming properties models used in computational cognitive neuroscience (e.g., information, representation, etc.) must literally exist system being modelled brain). Computational are important tools to test theory about how collected data behavioural or neuroimaging) has been generated. While usefulness unquestionable, it does not follow should entail construed model representation). this assumption present computationalist accounts, held across board neuroscience. In last section, offers dynamical with Dynamical Causal Modelling (DCM) combines systems (DST) mathematical formalisms theoretical contextualisation provided by Embodied and Enactive Cognitive Science (EECS).

Язык: Английский

---

Spatially-embedded recurrent neural networks reveal widespread links between structural and functional neuroscience findings

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2022, Номер unknown

Опубликована: Ноя. 18, 2022

ABSTRACT Brain networks exist within the confines of resource limitations. As a result, brain network must overcome metabolic costs growing and sustaining its physical space, while simultaneously implementing required information processing. To observe effect these processes, we introduce spatially-embedded recurrent neural (seRNN). seRNNs learn basic task-related inferences existing 3D Euclidean where communication constituent neurons is constrained by sparse connectome. We find that seRNNs, similar to primate cerebral cortices, naturally converge on solving using modular small-world networks, in which functionally units spatially configure themselves utilize an energetically-efficient mixed-selective code. all features emerge unison, reveal how many common structural functional motifs are strongly intertwined can be attributed biological optimization processes. serve as model systems bridge between research communities move neuroscientific understanding forward.

Язык: Английский

---

From abstract networks to biological realities

Physics of Life Reviews, Год журнала: 2024, Номер 49, С. 12 - 14

Опубликована: Фев. 29, 2024

Язык: Английский

---

Brain topology improved spiking neural network for efficient reinforcement learning of continuous control

Frontiers in Neuroscience, Год журнала: 2024, Номер 18

Опубликована: Апрель 16, 2024

The brain topology highly reflects the complex cognitive functions of biological after million-years evolution. Learning from these topologies is a smarter and easier way to achieve brain-like intelligence with features efficiency, robustness, flexibility. Here we proposed topology-improved spiking neural network (BT-SNN) for efficient reinforcement learning. First, hundreds are generated selected as subsets Allen mouse help Tanimoto hierarchical clustering algorithm, which has been widely used in analyzing key connectome. Second, few constraints filter out three candidates, including but not limited proportion node (e.g., sensation, memory, motor types) sparsity. Third, integrated hybrid numerical solver-improved leaky-integrated fire neurons. Fourth, algorithm then tuned an evolutionary named adaptive random search instead backpropagation guide synaptic modifications without affecting raw topology. Fifth, under test four animal-survival-like RL tasks (i.e., dynamic controlling Mujoco), BT-SNN can higher scores than only counterpart SNN using also some classical ANNs long-short-term memory multi-layer perception). This result indicates that research effort incorporating learning rules much store future.

Язык: Английский

---

Complete and partial synchronization in empirical brain networks

AEU - International Journal of Electronics and Communications, Год журнала: 2023, Номер 170, С. 154863 - 154863

Опубликована: Авг. 22, 2023

Язык: Английский

---

Causal Influences Decouple From Their Underlying Network Structure In Echo State Networks

2022 International Joint Conference on Neural Networks (IJCNN), Год журнала: 2022, Номер unknown, С. 1 - 8

Опубликована: Июль 18, 2022

Echo State Networks (ESN) are versatile recurrent neural network models in which the hidden layer remains unaltered during training. Interactions among nodes of this static backbone (the structure) produce diverse representations (i.e., dynamics) given stimuli that harnessed by a read-out mechanism to perform computations needed for solving task behavior). Moreover, ESNs accessible neuronal circuits, since they relatively inexpensive train. Therefore, have become attractive neuroscientists studying relationship between structure, function, and behavior. For instance, it is not yet clear how distinctive connectivity patterns brain networks (structure) support effective interactions their (dynamics) these give rise computation (behavior). To address question, we employed an ESN with biologically inspired structure used systematic multi-site lesioning framework quantify causal contribution each node network's output, thus providing link We then focused on structure-function decomposed influence all other nodes, using same framework. found properly engineered interact largely irrespective underlying structure. However, topology where ESN's leakage rate non-optimal dynamics diminished, determine interactions. Our results suggest relations can be into two components, direct indirect The former based influences relying structural connections. latter describe communication any through intermediate nodes. These widely distributed may crucially contribute efficient performance ESNs.

Язык: Английский

---

How to incorporate biological insights into network models and why it matters

The Journal of Physiology, Год журнала: 2022, Номер 601(15), С. 3037 - 3053

Опубликована: Сен. 25, 2022

Due to the staggering complexity of brain and its neural circuitry, neuroscientists rely on analysis mathematical models elucidate function. From Hodgkin Huxley's detailed description action potential in 1952 today, new theories increasing computational power have opened up novel avenues study how circuits implement computations that underlie behaviour. Computational developed many differ complexity, biological realism or emergent network properties. With recent advances experimental techniques for anatomical reconstructions large-scale activity recordings, rich data become more available. The challenge when building is reflect results, either through a high level detail by finding an appropriate abstraction. Meanwhile, machine learning has facilitated development artificial networks, which are trained perform specific tasks. While they proven successful at achieving task-oriented behaviour, often abstract constructs features from physiology circuits. Thus, it unclear whether mechanisms underlying computation can be investigated analysing networks accomplish same function but their mechanisms. Here, we argue biologically realistic crucial establishing causal relationships between neurons, synapses, More specifically, advocate consider connectivity structure recorded dynamics while evaluating task performance.

Язык: Английский

---

Biologically plausible models of cognitive flexibility: merging recurrent neural networks with full-brain dynamics

Current Opinion in Behavioral Sciences, Год журнала: 2024, Номер 56, С. 101351 - 101351

Опубликована: Фев. 6, 2024

Cognitive flexibility, a cornerstone of human cognition, enables us to adapt shifting environmental demands. This brain function has been widely explored using computational modeling, although oftentimes these models focus on the operational dimension cognitive flexibility and do not retain sufficient level neurobiological detail lead electrophysiological or neuroimaging insights. In this review, we explore recent advances future directions neurobiologically plausible flexibility. We first cover progress in recurrent neural network trained perform flexible tasks, followed by discussion how whole-brain large-scale have approached distributed nature functions. Ultimately, propose here hybrid framework which both modeling philosophies converge, advocating for balanced approach that merges power with realistic spatiotemporal dynamics activity, early examples direction.

Язык: Английский

---

Unlocking the Future of Drug Development: Generative AI, Digital Twins, and Beyond

Опубликована: Март 15, 2024

This article delves into the intersection of generative AI and digital twins within drug discovery, exploring their synergistic potential to revolutionize pharmaceutical research development. Through various instances examples, we illuminate how algorithms, capable simulating vast chemical spaces predicting molecular properties, are increasingly integrated with biological systems expedite discovery. By harnessing power computational models machine learning, researchers can design novel compounds tailored specific targets, optimize candidates, simulate behavior virtual environments. paradigm shift offers unprecedented opportunities for accelerating development, reducing costs, and, ultimately, improving patient outcomes. As navigate this rapidly evolving landscape, collaboration between interdisciplinary teams continued innovation will be paramount in realizing promise advancing

Язык: Английский

---