Current State and Future Directions for Learning in Biological Recurrent Neural Networks: A Perspective Piece

Prince, Luke Y.; Eyono, Roy Henha; Boven, Ellen; Ghosh, Arna; Joe, Pemberton,; Scherr, Franz; Clopath, Claudia; Costa, Rui Ponte; Maass, Wolfgang; Richards, Blake A.; Savin, Cristina; Wilmes, Katharina Anna

doi:10.48550/arxiv.2105.05382

Cited by 3 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Altogether, reservoir computing, and recurrent neural network models in general, open a world of possibilities to provide mechanistic explanations for the way computations take place in real brain networks. By avoiding biologically implausible credit assignments derived from brackpropagation training [34], reservoir computing becomes ideal for this purpose. The general conn2res workflow requires the following parameters to be provided by the user: i) a task name or a supervised learning dataset; ii) a connectome or connectivity matrix; iii) a set of input nodes; iv) a set of readout nodes or modules; v) the type of local dynamics, which can be either spiking neurons, artificial neurons (with a variety of activation functions), or memristive devices (for the simulation of physical reservoirs); and vi) the linear model to be trained in the readout module.…”

Section: The Brain As a Reservoirmentioning

confidence: 99%

“…In reservoir networks learning occurs only at the readout connections, and hence the main architecture of the reservoir does not require specific weight calibration, remaining fixed throughout training. This eliminates a confounder while avoiding biologically implausible credit assignment problems such as the use of backpropagation training [34]. These reasons make reservoir Figure 1.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

conn2res: A toolbox for connectome-based reservoir computing

Mišić

Suárez

Mihalik³

et al. 2023

Preprint

View full text Add to dashboard Cite

The connection patterns of neural circuits form a complex network. How signaling in these circuits manifests as complex cognition and adaptive behaviour remains the central question in neuroscience. Concomitant advances in connectomics and artificial intelligence open fundamentally new opportunities to understand how connection patterns shape computational capacity in biological brain networks. Reservoir computing is a versatile paradigm that uses nonlinear dynamics of high-dimensional dynamical systems to perform computations and approximate cognitive functions. Here we present conn2res: an open-source Python toolbox for implementing biological neural networks as artificial neural networks. conn2res is modular, allowing arbitrary architectures and arbitrary dynamics to be imposed. The toolbox allows researchers to input connectomes reconstructed using multiple techniques, from tract tracing to noninvasive diffusion imaging, and to impose multiple dynamical systems, from simple spiking neurons to memristive dynamics. The versatility of the conn2res toolbox allows us to ask new questions at the confluence of neuroscience and artificial intelligence. By reconceptualizing function as computation, conn2res sets the stage for a more mechanistic understanding of structure-function relationships in brain networks.

show abstract

Section: The Brain As a Reservoirmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

conn2res: A toolbox for connectome-based reservoir computing

Mišić

Suárez

Mihalik³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…It is known that RNNs are of importance in understanding the brain (Prince et al, 2021). It also is known that RNNs are, in theory, Turing Complete (Chung and Siegelmann 2021), and therefore, each point of an RNN's weight-space potentially represents a different possible algorithm.…”

Section: Introductionmentioning

confidence: 99%

A Minimal “Functionally Sentient” Organism Trained With Backpropagation Through Time

Pisheh Var,

Fairbank,

Samothrakis

2023

Adaptive Behavior

View full text Add to dashboard Cite

This article presents a scenario where a simple simulated organism must explore and exploit an environment containing a food pile. The organism learns to make observations of the environment, use memory to record those observations, and thus plan and navigate to the regions with the strongest food density. We compare different reinforcement learning algorithms with an adaptive dynamic programming algorithm and conclude that backpropagation through time can convincingly solve this recurrent neural-network challenge. Furthermore, we argue that this algorithm successfully mimics a minimal ‘functionally sentient’ organism’s fundamental objectives and mental environmental-mapping skills while seeking a food pile distributed statically or randomly in an environment.

show abstract

Connectome-based reservoir computing with the conn2res toolbox

Suárez,

Mihalik,

Milisav

et al. 2024

Nat Commun

View full text Add to dashboard Cite

The connection patterns of neural circuits form a complex network. How signaling in these circuits manifests as complex cognition and adaptive behaviour remains the central question in neuroscience. Concomitant advances in connectomics and artificial intelligence open fundamentally new opportunities to understand how connection patterns shape computational capacity in biological brain networks. Reservoir computing is a versatile paradigm that uses high-dimensional, nonlinear dynamical systems to perform computations and approximate cognitive functions. Here we present : an open-source Python toolbox for implementing biological neural networks as artificial neural networks. is modular, allowing arbitrary network architecture and dynamics to be imposed. The toolbox allows researchers to input connectomes reconstructed using multiple techniques, from tract tracing to noninvasive diffusion imaging, and to impose multiple dynamical systems, from spiking neurons to memristive dynamics. The versatility of the toolbox allows us to ask new questions at the confluence of neuroscience and artificial intelligence. By reconceptualizing function as computation, sets the stage for a more mechanistic understanding of structure-function relationships in brain networks.

show abstract

Current State and Future Directions for Learning in Biological Recurrent Neural Networks: A Perspective Piece

Cited by 3 publications

References 0 publications

conn2res: A toolbox for connectome-based reservoir computing

conn2res: A toolbox for connectome-based reservoir computing

A Minimal “Functionally Sentient” Organism Trained With Backpropagation Through Time

Connectome-based reservoir computing with the conn2res toolbox

Contact Info

Product

Resources

About