Cognitive behaviors originate in the responses of neuronal populations. We have a reasonable understanding of how the activity of a single neuron can be related to a specific behavior. However, it is still unclear how more complex behaviors are inferred from the responses of neuronal populations. This is a particularly timely problem because multi-neuronal recording techniques have recently become increasingly available, simultaneously spurring advances in the analysis of neuronal population data. These developments highlight the challenges of combining theoretical and experimental approaches because both approaches have their unique set of constraints.
We first study how behaviors are encoded in the responses of neuronal populations. For this we investigate the statistical nature of the neuronal response distributions that correspond to a given behavior or sensory input. This allows us to create encoding models that describe the data by making empirically driven assumptions, and to model the putative mechanisms that optimally combine the responses from multiple neurons. We then build on these findings to compute decoding models that predict behavioral outcomes corresponding to given neuronal population activities, e.g. using Bayesian inference. These decoders combined with task-specific read-out rules are the cornerstones of our understanding of how the activity of neuronal populations mediates behavioral outcomes. Finally, Brain-machine interfaces (BMIs) provide a direct validation of the decoding models because they are a closed-loop implementation between neuronal responses and behavior. BMIs also deal with additional complexities such as motor feedback and real-time constraints, both of which being an integral part of how neuronal populations mediate behavior. Most importantly, BMIs have profound bearing on medical applications where they can help cure and restore lost function, e.g. tetraplegia.
This Research Topic presents a unified view of how behavior is inferred from neuronal population responses using computational models that are empirically driven. We will present key findings for each stage of the journey taking us from neuronal population activities to behavior. The scope of this research topic is to pinpoint computational models that fall within the constraints of experimental data collection, and to provide a toolbox for the engineer and clinician to help develop and improve BMIs.
Cognitive behaviors originate in the responses of neuronal populations. We have a reasonable understanding of how the activity of a single neuron can be related to a specific behavior. However, it is still unclear how more complex behaviors are inferred from the responses of neuronal populations. This is a particularly timely problem because multi-neuronal recording techniques have recently become increasingly available, simultaneously spurring advances in the analysis of neuronal population data. These developments highlight the challenges of combining theoretical and experimental approaches because both approaches have their unique set of constraints.
We first study how behaviors are encoded in the responses of neuronal populations. For this we investigate the statistical nature of the neuronal response distributions that correspond to a given behavior or sensory input. This allows us to create encoding models that describe the data by making empirically driven assumptions, and to model the putative mechanisms that optimally combine the responses from multiple neurons. We then build on these findings to compute decoding models that predict behavioral outcomes corresponding to given neuronal population activities, e.g. using Bayesian inference. These decoders combined with task-specific read-out rules are the cornerstones of our understanding of how the activity of neuronal populations mediates behavioral outcomes. Finally, Brain-machine interfaces (BMIs) provide a direct validation of the decoding models because they are a closed-loop implementation between neuronal responses and behavior. BMIs also deal with additional complexities such as motor feedback and real-time constraints, both of which being an integral part of how neuronal populations mediate behavior. Most importantly, BMIs have profound bearing on medical applications where they can help cure and restore lost function, e.g. tetraplegia.
This Research Topic presents a unified view of how behavior is inferred from neuronal population responses using computational models that are empirically driven. We will present key findings for each stage of the journey taking us from neuronal population activities to behavior. The scope of this research topic is to pinpoint computational models that fall within the constraints of experimental data collection, and to provide a toolbox for the engineer and clinician to help develop and improve BMIs.