Project

Back to overview

Prospective coding with pyramidal neurons

English title Prospective coding with pyramidal neurons
Applicant Senn Walter
Number 156863
Funding scheme Project funding (Div. I-III)
Research institution Institut für Physiologie Medizinische Fakultät Universität Bern
Institution of higher education University of Berne - BE
Main discipline Neurophysiology and Brain Research
Start/End 01.12.2015 - 30.11.2018
Approved amount 625'000.00
Show all

All Disciplines (2)

Discipline
Neurophysiology and Brain Research
Mathematics

Keywords (5)

pyramidal neurons; computational neuroscience; learning theory; synaptic plasticity; dendritic integration and spikes

Lay Summary (German)

Lead
Prospektive Kodierung in kortikalen Pyramidenzellen.Pyramidenzellen dominieren die Vernetzung und die Verarbeitung von Information im Gehirn. Mit neuen experimentellen Methoden lassen sich elektrische Ströme von Pyramidenzellen im Hirn der Maus messen, während dem das Tier zum Beispiel eine senso-motorische Assoziation lernt. Das Projekt untersucht die Hypothese, dass Pyramidenzellen sensorische Stimuli voraussagen und mit Hilfe des Voraussagefehlers die geplante motorische Antwort vorausschauend anpassen.
Lay summary

Etwa 80% der Nervenzellen im Grosshirn von Säugern bestehen aus Pyramidenzellen. Sie weisen eine bipolare dendritischen Struktur auf, die sich über verschiedene Schichten der Hirnrinde erstreckt. Wie diese Zellen allerdings sensorische und mentale Informationen verarbeiten und was die Rolle der auf- und absteigenden dendritischen Äste ist, bleibt unklar. Eine Hypothese besagt, dass die Nervenzellen sensorische Stimuli in den absteigenden Ästen und intern generierte Erwartungen in den aufsteigenden Ästen verarbeiten. Im Soma der Pyramidenzelle würden dann die beiden Informationsflüsse gegeneinander abgeglichen. Das Projekt untersucht mit theoretischen und experimentellen Methoden, wie solche Stimuli und Erwartungen innerhalb einer Nervenzelle kodiert werden und wie die Pyramidenzelle den Vergleich bewerkstelligen kann.

 

Ausgangspunkt des Projektes ist die neue Hypothese, dass Nervenzellen allgemein als Voraussage-Elemente dienen und nicht wie anhin angenommen, primär eine nichtlineare Recheneinheit darstellen. Diese Hypothese verleiht der einzelnen Nervenzellen eine Spur einer kognitiven Verarbeitung – nämlich aufgrund aktueller Stimuli und gemachten Erfahrungen die nächste Zukunft vorauszusagen. Falls die Nervenzelle als kleinste Verarbeitungseinheit im Gehirn bereits eine kognitive Komponente als Keim in sich trägt, so die Hoffnung, lassen sich in Netzwerken von Nervenzellen die kognitiven Aufgaben einfacher bewältigen.

 

Drei Forschungsgruppen sind in das Projekt involviert: Walter Senn (Universität Bern) arbeitet das mathematische Modell aus, Matthew Larkum (Humboldt Universität Berlin) testet die resultierenden Hypothesen an Verhaltensexperimenten in Mäusen und leitet gleichzeitig elektrische Signale von kortikalen Pyramidenzellen ab, und Thomas Nevian testet Hypothesen über die Plastizität von Synapsen an den auf- und absteigenden dendritischen Ästen, die dem Lernen einer Voraussage zugrunde liegen.

 

Direct link to Lay Summary Last update: 06.01.2015

Responsible applicant and co-applicants

Employees

Publications

Publication
Computational roles of plastic probabilistic synapses
Llera-Montero Milton, Sacramento João, Costa Rui Ponte (2019), Computational roles of plastic probabilistic synapses, in Current Opinion in Neurobiology, 54, 90-97.
Electrical Compartmentalization in Neurons
Wybo Willem A.M., Torben-Nielsen Benjamin, Nevian Thomas, Gewaltig Marc-Oliver (2019), Electrical Compartmentalization in Neurons, in Cell Reports, 26(7), 1759-1773.
A Perspective on Cortical Layering and Layer-Spanning Neuronal Elements
Larkum Matthew E., Petro Lucy S., Sachdev Robert N. S., Muckli Lars (2018), A Perspective on Cortical Layering and Layer-Spanning Neuronal Elements, in Frontiers in Neuroanatomy, 12, 1-9.
Dendritic cortical microcircuits approximate the backpropagation algorithm
Sacramento Joao, Ponte CostaRui, BengioYoshua, SennWalter (2018), Dendritic cortical microcircuits approximate the backpropagation algorithm, in Advances in Neural Information Processing Systems 31 (NIPS 2018), 1-9.
Dendritic calcium spikes are clearly detectable at the cortical surface
Suzuki Mototaka, Larkum Matthew E. (2017), Dendritic calcium spikes are clearly detectable at the cortical surface, in Nature Communications, 8(1), 276-276.
A Reward-Based Behavioral Platform to Measure Neural Activity during Head-Fixed Behavior
Micallef Andrew H., Takahashi Naoya, Larkum Matthew E., Palmer Lucy M. (2017), A Reward-Based Behavioral Platform to Measure Neural Activity during Head-Fixed Behavior, in Frontiers in Cellular Neuroscience, 11, 1-8.
Active cortical dendrites modulate perception
Takahashi Naoya, Oertner Thomas G., Hegemann Peter, Larkum Matthew E. (2016), Active cortical dendrites modulate perception, in Science, 354(6319), 1587-1590.
Prospective Coding by Spiking Neurons
Brea Johanni, Gaál Alexisz Tamás, Urbanczik Robert, Senn Walter (2016), Prospective Coding by Spiking Neurons, in PLOS Computational Biology, 12(6), 1-25.
Somato-dendritic Synaptic Plasticity and Error-backpropagation in Active Dendrites
Schiess Mathieu, Urbanczik Robert, Senn Walter (2016), Somato-dendritic Synaptic Plasticity and Error-backpropagation in Active Dendrites, in PLOS Computational Biology, 12(2), 1-18.
Hierarchical Novelty-Familiarity Representation in the Visual System by Modular Predictive Coding
Vladimirskiy Boris, Urbanczik Robert, Senn Walter (2015), Hierarchical Novelty-Familiarity Representation in the Visual System by Modular Predictive Coding, in PLOS ONE, 10(12), 1-19.

Collaboration

Group / person Country
Types of collaboration
Yoshua Bengio Canada (North America)
- in-depth/constructive exchanges on approaches, methods or results

Communication with the public

Communication Title Media Place Year
Talks/events/exhibitions Cosyne 2016 (Utah, selected plenary oral talk, audience ~3'000): Bayesian cue combination by International 2016

Associated projects

Number Title Start Funding scheme
180316 Neural Processing of Distinct Prediction Errors: Theory, Mechanisms & Interventions 01.09.2018 Sinergia
133094 Dendritic pointers and time multiplexing as cortical binding mechanisms 01.05.2011 Project funding (Div. I-III)
133094 Dendritic pointers and time multiplexing as cortical binding mechanisms 01.05.2011 Project funding (Div. I-III)
147636 Learning from delayed and sparse feedback 01.12.2013 Sinergia

Abstract

This Lead Agency proposal is a continuation of the personal SNF-grant of W. Senn on the theory of dendritic computation. In the running project period a key insight has been achieved by suggesting that learning on the level of a neuron implies the prediction of somatic spiking by the dendritic input (Urbanczik & Senn, Neuron 2014). This hypothesis introduces a paradigm shift in viewing dendritic computation and opens the door for a putative computational understanding of the signal processing in pyramidal neurons. So far, general experimental and theoretical research has tried to prove more and more complex nonlinearities in the dendritic processing of synaptic signals. But the mere description of a neuron as a complex input-output element gives only little insight into what dendrites are actually computing. In contrast, regarding neurons, and in particular pyramidal neurons, as intrinsic prediction elements links single neuron processing to a possible broader computational task.The current proposal extends this single neuron hypothesis by the notion of prospective coding. This notion introduces the idea that the activity of a neuron not only predicts current, but also future synaptic inputs. The proposal takes account of the bipolar dendritic morphology of a pyramidal neuron with a basal and apical dendritic tree. We hypothesize that both the basal and apical tree make independent predictions of the somatic spiking. These predictions are based on within-network projections to the basal tree, and extrinsic or top-down connections to the apical tree. The match between the independent predictions represents a high confidence signal that generates a dendritic calcium spike with a subsequent burst of somatic action potentials. These bursts can then be fed back to the presynaptic neurons that can use them as a teaching signal for their own up-stream synapses. The theory and its experimental verification is divided into 4 subprojects:SP1: Prospective coding (Lead: Senn lab, 1 PhD). Formalize the concept of prospective coding and show that the independent prediction of future input by the basal and dendritic trees is equivalent to a Bayesian cue combination problem.SP2: Backpropagation in time (Lead: Senn lab, 1 postdoc). Show that the matching signal for the prediction of future events can be used to train hidden neurons that contribute to these predictions. Apply the theory to the non-Markovian sequence learning problem and to a simplified ball catching problem.SP3: Novelty coding (Lead: Larkum lab, 1 postdoc - DFG). Test in vivo whether a dendritic calcium spike is representing the match between prediction signals or the match between novelty signals generated by the basal and apical trees. Verify the prediction of the cue combination hypothesis by measuring the neuronal responses to a somato-sensory oddball paradigm with combined auditory and somato-sensory cues.SP4: Error-correcting plasticity (Lead: Nevian lab, 1 PhD). Verify the hypothesis in vitro whether synaptic plasticity both in excitatory and inhibitory plasticity is error-correcting and hence non-Hebbian. Test whether plasticity involving calcium spikes has a longer induction time window as predicted by prospective coding.The first two subprojects are yield the formal framework in which the subsequent two experimental subprojects are embedded. The experiments are formulated such that, ideally, they verify or falsify the theory inspired hypotheses. They will be jointly designed and the results will be described in terms of a mathematical model.
-