Informationsverarbeitung natürlicher Stimuli im auditorischen System von Säugetieren

• Applicant: Dr. Christian Machens
• Subject Area: Cognitive, Systems and Behavioural Neurobiology
• Term: from 2006 to 2013
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 18059970

• Final Report Year: 2015

Final Report Abstract

A lot of previous research has elucidated how information is represented and processed in single neurons in the brain. However, information is often distributed across large populations of neurons. In this project, we studied exactly how information is represented in neural populations, and how neurons interact with each other in order to coordinate the precise distribution of information. We studied these questions in a range of different species and brain areas, including systems that process sensory information (such as the visual system of the fly or the somatosensory system of the monkey) and systems that underlie decision-making (such as the prefrontal cortices in rodents and monkeys). A common observation is that single neurons in higher-order brain regions mix or confuse various types of information. In other words, these neurons will represent sensory information, they will represent information about past events (“short-term memory”), and information about impending actions or movements (“decisions”). Since all action potentials fired by neurons are equal, it is not immediatly clear how other neurons (or the experimental observer) can distinguish the various types of information. By developing new methods for analyzing these recordings, we could show that despite the confusion of information at the single neuron level, the relevant information can be perfectly retrieved (demixed) at the level of the (complete) neural population. Our work shows that the messiness at the level of single neurons gives rise to a suprisingly clear organization of information representation at the level of whole networks of neurons. How does this precise distribution of information come about? Neurons interact with each other, but they somehow need to coordinate their interaction in order to generate the population representation that we observed. Starting from first principles, we showed that the efficient sharing of information among neurons requires rapid exchange of information within the network so that no two neurons represent exactly the same information. This rapid information exchange can be mediated by fast inhibition between neurons. We could show that the resulting networks share many of the biological features often observed in real networks such as the precise balance of excitatory and inhibitory inputs into each neuron, the irreliability of neural firing at the level of single neurons, etc. Furthermore, our framework predicts that neural networks compensate any perturbative change, be it the loss of neurons, or any other perturbation, almost instantaneously. The theory therefore offers a possible explanation for the astounding robustness of the brain against (limited) tissue damage, be it from experimental lesions or neural diseases.

Publications

• (2007). From response to stimulus: adaptive sampling in sensory physiology. Curr Opin Neurobiol 17(4):430–6
  Benda J, Gollisch T, Machens CK, Herz AVM
  
• (2010). Functional, but not anatomical separation of “what” and “when” in prefrontal cortex. J Neurosci, 30(1):350–60
  Machens CK, Romo R, Brody CD
  
• (2010). Spatio-temporal response properties of optic-flow processing neurons. Neuron, 67:628–41
  Weber F, Machens CK, Borst A
  
• (2011). Demixed principal component analysis. In: Advances in Neural Information Processing Systems 24 (epub)
  Brendel W, Romo R, Machens CK
  
• (2012). Disentangling the functional consequences of the connectivity between optic-flow processing neurons. Nat Neurosci, 15(3):441–8
  Weber F, Machens CK, Borst A
  (See online at https://doi.org/10.1038/nn.3044)
• (2013). Firing rate calculations in optimal spiking networks. In: Advances in Neural Information Processing Systems 26 (epub)
  Barrett DGT, Denève S, Machens CK
  
• (2013). Population-wide distributions of neural activity during perceptual decision-making. Prog Neurobiol 103:156–93
  Wohrer A, Humphries M, Machens CK
  
• (2014). Optogenetic perturbations reveal the dynamics of an oculomotor integrator. Front Neural Circuits 8:10
  Gonçalves P, Arrenberg A, Hablitzel B, Baier H, Machens CK
  (See online at https://doi.org/10.3389/fncir.2014.00010)
• (2015). On the number of neurons and time scale of integration underlying the formation of percepts in the brain. PLoS Comput Biol 11(3):e1004082
  Wohrer A, Machens CK
  (See online at https://doi.org/10.1371/journal.pcbi.1004082)
• Optimal compensation for neuron death. eLife Bd. 5 e12454 (PDF: 63 S.) Dec 2016
  Barrett DGT, Deneve S, Machens CK