one thing nice about statistics is that you can be useful in a wide variety of areas. The surprising thing is that the methods used are similar in both areas. -----------------------------------
The right framework for such a theory appears to be stochastic optimal control on the motor side and Bayesian inference on the sensory side; the two are dual in a deep mathematical sense.
Statistically, this involves stochastic process models such as non-homogeneous Poisson processes, inference using simulation-based methods including Markov chain Monte Carlo and the Bootstrap, and ideas from the emerging field of functional data analysis. Scientifically, one of the fundamental problems, which might be viewed as central to the enterprise, is to determine how neurons carry information. An exciting challenge is to make good use of the large amounts of data produced from many simultaneously-recorded neurons.
Optimal Bayesian Estimator: Such an estimator takes into account sensory data,recent control signals, knowledge of body dynamics, as well as its earlier output, and weights all these sources of information regarding the current state in proportion to their reliability. In modeling practice one typically uses a Kalman filter—which is the optimal estimator when the dynamics and sensory measurements are linear and the noise is Gaussian, and provides a good approximation in other cases.A number of studies suggest that perception in general, and online state estimation in particular, are based on the principles of Bayesian inference. A key feature of optimal estimators is their ability to anticipate state changes before the corresponding sensory data have arrived. This requires either explicit or implicit knowledge of body dynamics, that is, an ‘internal model’.
Competition among parallel simultaneous representations of potential actions:
When there are multiple graspable objects (distractors) in the reach and one needs to decide (intention) which one to grasp, how does the decoder respond? This brings us again to the problem of fusing information from multiple regions; where decisions are made and plannings of how to execute the grasp are done.
8 comments:
one thing nice about statistics is that you can be useful in a wide variety of areas.
The surprising thing is that the methods used are similar in both areas.
-----------------------------------
The right framework for such a theory appears to be stochastic optimal control on the motor side and Bayesian inference on the sensory side; the two are dual in a deep mathematical sense.
Wiki article
http://en.wikipedia.org/wiki/Brain-computer_interface
I love the interdisciplinary aspect of this research. I was attracted by both the mathematics and the neuroscience in it.
Statistically, this involves stochastic process models such as non-homogeneous Poisson processes, inference using simulation-based methods including Markov chain Monte Carlo and the Bootstrap, and ideas from the emerging field of functional data analysis. Scientifically, one of the fundamental problems, which might be viewed as central to the enterprise, is to determine how neurons carry information. An exciting challenge is to make good use of the large amounts of data produced from many simultaneously-recorded neurons.
Optimal Bayesian Estimator:
Such an estimator takes into account sensory data,recent control signals, knowledge of body dynamics, as well as its earlier
output, and weights all these sources of information regarding the current state in proportion to their reliability. In modeling practice one typically uses a Kalman filter—which is the optimal estimator when the dynamics and sensory measurements are linear and the noise is Gaussian, and provides a good approximation in other cases.A number of studies suggest that perception in general, and online state estimation in particular, are based on the principles of Bayesian
inference. A key feature of optimal estimators is their ability to anticipate state changes before the corresponding sensory data have arrived. This requires either explicit or implicit knowledge of body dynamics, that is, an ‘internal model’.
The use of multiple arrays will allow us to explor neural control using activity from multiple brain areas (parietal in addition to motor cortex).
Competition among parallel simultaneous representations of potential actions:
When there are multiple graspable objects (distractors) in the reach and one needs to decide (intention) which one to grasp, how does the decoder respond? This brings us again to the problem of fusing information from multiple regions; where decisions are made and plannings of how to execute the grasp are done.
Post a Comment