Machine Learning for Neural Decoding

eNeuro, Journal Year: 2020, Volume and Issue: 7(4), P. ENEURO.0506 - 19.2020

Published: July 1, 2020

Abstract Despite rapid advances in machine learning tools, the majority of neural decoding approaches still use traditional methods. Modern which are versatile and easy to use, have potential significantly improve performance. This tutorial describes how effectively apply these algorithms for typical problems. We provide descriptions, best practices, code applying common methods, including networks gradient boosting. also detailed comparisons performance various methods at task spiking activity motor cortex, somatosensory hippocampus. particularly ensembles, outperform approaches, such as Wiener Kalman filters. Improving allows neuroscientists better understand information contained a population can help advance engineering applications brain–machine interfaces. Our package is available github.com/kordinglab/neural_decoding .

Language: Английский

---

The mouse prefrontal cortex: Unity in diversity

Neuron, Journal Year: 2021, Volume and Issue: 109(12), P. 1925 - 1944

Published: April 23, 2021

Language: Английский

---

The Computational and Neural Bases of Context-Dependent Learning

Annual Review of Neuroscience, Journal Year: 2023, Volume and Issue: 46(1), P. 233 - 258

Published: March 27, 2023

Flexible behavior requires the creation, updating, and expression of memories to depend on context. While neural underpinnings each these processes have been intensively studied, recent advances in computational modeling revealed a key challenge context-dependent learning that had largely ignored previously: Under naturalistic conditions, context is typically uncertain, necessitating contextual inference. We review theoretical approach formalizing face uncertainty core computations it requires. show how this begins organize large body disparate experimental observations, from multiple levels brain organization (including circuits, systems, behavior) regions (most prominently prefrontal cortex, hippocampus, motor cortices), into coherent framework. argue inference may also be understanding continual brain. This theory-driven perspective places as component learning.

Language: Английский

---

Sensory coding and the causal impact of mouse cortex in a visual decision

eLife, Journal Year: 2021, Volume and Issue: 10

Published: July 30, 2021

Correlates of sensory stimuli and motor actions are found in multiple cortical areas, but such correlates do not indicate whether these areas causally relevant to task performance. We trained mice discriminate visual contrast report their decision by steering a wheel. Widefield calcium imaging Neuropixels recordings cortex revealed stimulus-related activity (VIS) frontal (MOs) widespread movement-related across the whole dorsal cortex. Optogenetic inactivation biased choices only when targeted at VIS MOs,proportionally each site's encoding stimulus, times corresponding peak stimulus decoding. A neurometric model based on summing subtracting MOs successfully described behavioral performance predicted effect optogenetic inactivation. Thus, signals localized play causal role performance, while correlating with movement reflect processes that role.

Language: Английский

---

50 years of mnemonic persistent activity: quo vadis?

Trends in Neurosciences, Journal Year: 2021, Volume and Issue: 44(11), P. 888 - 902

Published: Oct. 13, 2021

Language: Английский

---

Neural activity in the mouse claustrum in a cross-modal sensory selection task

Neuron, Journal Year: 2021, Volume and Issue: 110(3), P. 486 - 501.e7

Published: Dec. 3, 2021

Language: Английский

---

Modularity and robustness of frontal cortical networks

Cell, Journal Year: 2021, Volume and Issue: 184(14), P. 3717 - 3730.e24

Published: July 1, 2021

Language: Английский

---

Neural Algorithms and Circuits for Motor Planning

Annual Review of Neuroscience, Journal Year: 2022, Volume and Issue: 45(1), P. 249 - 271

Published: March 22, 2022

The brain plans and executes volitional movements. underlying patterns of neural population activity have been explored in the context movements eyes, limbs, tongue, head nonhuman primates rodents. How do networks neurons produce slow dynamics that prepare specific fast ultimately initiate these movements? Recent work exploits rapid calibrated perturbations to test dynamical systems models are capable producing observed activity. These joint experimental computational studies show cortical during motor planning reflect fixed points (attractors). Subcortical control signals reshape move attractors over multiple timescales, causing commitment actions transitions movement execution. Experiments rodents beginning reveal how algorithms implemented at level brain-wide circuits.

Language: Английский

---

Cortical feedback loops bind distributed representations of working memory

Nature, Journal Year: 2022, Volume and Issue: 608(7922), P. 381 - 389

Published: July 27, 2022

Working memory-the brain's ability to internalize information and use it flexibly guide behaviour-is an essential component of cognition. Although activity related working memory has been observed in several brain regions1-3, how neural populations actually represent memory4-7 the mechanisms by which this is maintained8-12 remain unclear13-15. Here we describe implementation visual mice alternating between a delayed non-match-to-sample task simple discrimination that does not require but identical stimulus, movement reward statistics. Transient optogenetic inactivations revealed distributed areas neocortex were required selectively for maintenance memory. Population area AM premotor M2 during delay period was dominated orderly low-dimensional dynamics16,17 were, however, independent Instead, representations embedded high-dimensional population activity, present both cortical areas, persisted throughout inter-stimulus period, predicted behavioural responses task. To test whether nature dependent on reciprocal interactions regions18-20, silenced one (AM or M2) while recording feedback received from other. inactivation either led selective disruption inter-areal communication Therefore, reciprocally interconnected maintain bound

Language: Английский

---

Thalamus-driven functional populations in frontal cortex support decision-making

Nature Neuroscience, Journal Year: 2022, Volume and Issue: 25(10), P. 1339 - 1352

Published: Sept. 28, 2022

Neurons in frontal cortex exhibit diverse selectivity representing sensory, motor and cognitive variables during decision-making. The neural circuit basis for this complex remains unclear. We examined activity mediating a tactile decision mouse anterior lateral relation to the underlying circuits. Contrary notion of randomly mixed selectivity, an analysis 20,000 neurons revealed organized coding behavior. Individual exhibited prototypical response profiles that were repeatable across mice. Stimulus, choice action coded nonrandomly by distinct neuronal populations could be delineated their profiles. related long-range inputs from somatosensory cortex, contralateral thalamus. Each input connects all functional but with differing strength. Task was more strongly dependent on thalamic than cortico-cortical inputs. Our results suggest thalamus drives subnetworks within features

Language: Английский

---