We gratefully acknowledge support from
the Simons Foundation and member institutions.

Neurons and Cognition

New submissions

[ total of 7 entries: 1-7 ]
[ showing up to 2000 entries per page: fewer | more ]

New submissions for Wed, 25 Nov 20

[1]  arXiv:2011.11710 [pdf, other]
Title: Natural-gradient learning for spiking neurons
Comments: Joint senior authorship: Walter M. Senn and Mihai A. Petrovici
Subjects: Neurons and Cognition (q-bio.NC); Neural and Evolutionary Computing (cs.NE); Differential Geometry (math.DG); Computation (stat.CO)

In many normative theories of synaptic plasticity, weight updates implicitly depend on the chosen parametrization of the weights. This problem relates, for example, to neuronal morphology: synapses which are functionally equivalent in terms of their impact on somatic firing can differ substantially in spine size due to their different positions along the dendritic tree. Classical theories based on Euclidean gradient descent can easily lead to inconsistencies due to such parametrization dependence. The issues are solved in the framework of Riemannian geometry, in which we propose that plasticity instead follows natural gradient descent. Under this hypothesis, we derive a synaptic learning rule for spiking neurons that couples functional efficiency with the explanation of several well-documented biological phenomena such as dendritic democracy, multiplicative scaling and heterosynaptic plasticity. We therefore suggest that in its search for functional synaptic plasticity, evolution might have come up with its own version of natural gradient descent.

[2]  arXiv:2011.12050 [pdf, ps, other]
Title: Role of inhibitory neurons in temporal correlations of critical and supercritical spontaneous activity
Comments: 15 pages, 6 figures
Journal-ref: Physica A, 125555 (2020)
Subjects: Neurons and Cognition (q-bio.NC); Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech)

Experimental and numerical results suggest that the brain can be viewed as a system acting close to a critical point, as confirmed by scale-free distributions of relevant quantities in a variety of different systems and models. Less attention has received the investigation of the temporal correlation functions in brain activity in different, healthy and pathological, conditions. Here we perform this analysis by means of a model with short and long-term plasticity which implements the novel feature of different recovery rates for excitatory and inhibitory neurons, found experimentally. We evidence the important role played by inhibitory neurons in the supercritical state: We detect an unexpected oscillatory behaviour of the correlation decay, whose frequency depends on the fraction of inhibitory neurons and their connectivity degree. This behaviour can be rationalized by the observation that bursts in activity become more frequent and with a smaller amplitude as inhibition becomes more relevant.

[3]  arXiv:2011.12067 [pdf, other]
Title: Dendritic trafficking: synaptic scaling and structural plasticity
Comments: Submitted to the European Control Conference 2021
Subjects: Neurons and Cognition (q-bio.NC); Systems and Control (eess.SY)

Neuronal circuits internally regulate electrical signaling via a host of homeostatic mechanisms. Two prominent mechanisms, synaptic scaling and structural plasticity, are believed to maintain average activity within an operating range by modifying the strength and spatial extent of network connectivity using negative feedback. However, both mechanisms operate on relatively slow timescales and thus face fundamental limits due to delays. We show that these mechanisms fulfill complementary roles in maintaining stability in a large network. In particular, even relatively, slow growth dynamics improves performance significantly beyond synaptic scaling alone.

Cross-lists for Wed, 25 Nov 20

[4]  arXiv:2011.12012 (cross-list from cs.NE) [pdf, other]
Title: A More Biologically Plausible Local Learning Rule for ANNs
Comments: 8 pages (4 main + 1 reference + 3 supplementary)
Subjects: Neural and Evolutionary Computing (cs.NE); Neurons and Cognition (q-bio.NC)

The backpropagation algorithm is often debated for its biological plausibility. However, various learning methods for neural architecture have been proposed in search of more biologically plausible learning. Most of them have tried to solve the "weight transport problem" and try to propagate errors backward in the architecture via some alternative methods. In this work, we investigated a slightly different approach that uses only the local information which captures spike timing information with no propagation of errors. The proposed learning rule is derived from the concepts of spike timing dependant plasticity and neuronal association. A preliminary evaluation done on the binary classification of MNIST and IRIS datasets with two hidden layers shows comparable performance with backpropagation. The model learned using this method also shows a possibility of better adversarial robustness against the FGSM attack compared to the model learned through backpropagation of cross-entropy loss. The local nature of learning gives a possibility of large scale distributed and parallel learning in the network. And finally, the proposed method is a more biologically sound method that can probably help in understanding how biological neurons learn different abstractions.

Replacements for Wed, 25 Nov 20

[5]  arXiv:2007.05395 (replaced) [pdf]
Title: On the variability of functional connectivity and network measures in source-reconstructed EEG time-series
Subjects: Neurons and Cognition (q-bio.NC); Signal Processing (eess.SP)
[6]  arXiv:1911.08241 (replaced) [pdf, other]
Title: Complex Sparse Code Priors Improve the Statistical Models of Neurons in Primate Primary Visual Cortex
Comments: 13 pages, 10 figures
Subjects: Quantitative Methods (q-bio.QM); Neurons and Cognition (q-bio.NC)
[7]  arXiv:2002.11006 (replaced) [pdf, ps, other]
Title: Transient chaotic dimensionality expansion by recurrent networks
Comments: Fully overhauled version with lots of new material
Subjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Neurons and Cognition (q-bio.NC)
[ total of 7 entries: 1-7 ]
[ showing up to 2000 entries per page: fewer | more ]

Disable MathJax (What is MathJax?)

Links to: arXiv, form interface, find, q-bio, recent, 2011, contact, help  (Access key information)