Difference between revisions of "Main Page"

From Sean_Carver
Jump to: navigation, search
Line 19: Line 19:
 
'''Research Projects:'''  One of the challenges of neural system identification is the reality that the neural systems generating the experimental data remain inevitably more complicated than the models being fitted.  What sort of problems this presents in practice is unknown.  For final projects, students will test approximate maximum likelihood parameter estimation under the situation just described (the generating system is more complicated than the fitted model).  To perform these tests students will choose between a one, two, or three compartment model to generate the data, and will fit one of these three model structures to the simulated data.  To generate the data, students will decide what injected current stimulus to provide to each compartment.  To fit the data, students will specify the correctly known, incorrectly assumed, and unknown parameters.  Students will make all choices by editing a single MATLAB script. It will take 5-15 minutes to do the editing needed to run an experiment and perhaps overnight to run the experiment.  Mid-semester, students will submit a proposal for their project, and, thereafter, run one experiment between each class.
 
'''Research Projects:'''  One of the challenges of neural system identification is the reality that the neural systems generating the experimental data remain inevitably more complicated than the models being fitted.  What sort of problems this presents in practice is unknown.  For final projects, students will test approximate maximum likelihood parameter estimation under the situation just described (the generating system is more complicated than the fitted model).  To perform these tests students will choose between a one, two, or three compartment model to generate the data, and will fit one of these three model structures to the simulated data.  To generate the data, students will decide what injected current stimulus to provide to each compartment.  To fit the data, students will specify the correctly known, incorrectly assumed, and unknown parameters.  Students will make all choices by editing a single MATLAB script. It will take 5-15 minutes to do the editing needed to run an experiment and perhaps overnight to run the experiment.  Mid-semester, students will submit a proposal for their project, and, thereafter, run one experiment between each class.
  
'''Questions?''' Email the instructor at ''sean'' [dot] ''carver'' [at] ''jhu'' [dot] ''edu''.  Alternatively, I will be presenting a poster about this class at the Annual Meeting of the Society of Neuroscience.  You can come talk to me during the Department of Psychological and Brain Sciences practice poster session.  I will update this page when I find out exactly when and where this poster session will be.  (Nov 12, I think).
+
'''Questions?''' Email the instructor at ''sean'' [dot] ''carver'' [at] ''jhu'' [dot] ''edu''.  Alternatively, I will be presenting a poster about this class at the Annual Meeting of the Society of Neuroscience.  You can come talk to me during the Department of Psychological and Brain Sciences practice poster session.  this poster session will be, scheduled for November 12 from 12:00 - 1:00 in 217 Ames.
  
 
'''Note:''' My homepage has moved to the [http://limbs.lcsr.jhu.edu/User:Scarver LIMBS wiki].
 
'''Note:''' My homepage has moved to the [http://limbs.lcsr.jhu.edu/User:Scarver LIMBS wiki].

Revision as of 20:58, 11 November 2008

Modeling and Identifying Neural Systems

Instructor: Sean G. Carver, Ph.D., Psychological and Brain Sciences, The Johns Hopkins University.

Semester Offered: Spring 2009.

One Hundred Word Description: Students in this course will be trained to perform original research in computational neuroscience. The course will cover mathematical modeling of neurons, which is useful for understanding the computations of single cells. The student's research will test software, adapted by the instructor from methods of other disciplines, for systematically creating models of neurons using experimental data. For the tests, data will come from another known model, rather than from a biological neuron. To perform the research, students will be given a thorough understanding of the biophysical mechanisms of neurons and of the basic paradigms of neural modeling and system identification.

Prerequisites: Calculus I and II, and Nervous System I and II, or permission of instructor.

Background: Neural modeling is often pursued in an ad hoc way. Researchers add the mechanisms they know about, but need to wave their hands about the ones they don't. They necessarily make many simplifying assumptions but often include many details that are not needed to parsimoniously capture the phenomena. More...

Course Mechanics: This class will be a hands on experience. Pending approval, class will meet twice a week in the Kreiger computer classroom. Each meeting will last 1.5 hours. In addition, there will be three hours per week of supervised computer laboratory time. Attendance for the laboratory will be optional. The purpose of the laboratory will be to allow students, if they choose, to complete computer assignments with the help of the instructor. An effort will be made to design the homework sets to allow most students to complete most of the assignments during the laboratory time. Grading will be determined partly by homework (due at the beginning of most class periods) and partly by final projects. I am not planning exams.

Textbook: Neurons in Action: Tutorials and Simulations Using NEURON, Version 2. By John W. Moore and Ann E. Stuart. Sinauer Associates, Inc, Sunderland Massachusetts.

Tentative Syllabus: (click here)

Research Projects: One of the challenges of neural system identification is the reality that the neural systems generating the experimental data remain inevitably more complicated than the models being fitted. What sort of problems this presents in practice is unknown. For final projects, students will test approximate maximum likelihood parameter estimation under the situation just described (the generating system is more complicated than the fitted model). To perform these tests students will choose between a one, two, or three compartment model to generate the data, and will fit one of these three model structures to the simulated data. To generate the data, students will decide what injected current stimulus to provide to each compartment. To fit the data, students will specify the correctly known, incorrectly assumed, and unknown parameters. Students will make all choices by editing a single MATLAB script. It will take 5-15 minutes to do the editing needed to run an experiment and perhaps overnight to run the experiment. Mid-semester, students will submit a proposal for their project, and, thereafter, run one experiment between each class.

Questions? Email the instructor at sean [dot] carver [at] jhu [dot] edu. Alternatively, I will be presenting a poster about this class at the Annual Meeting of the Society of Neuroscience. You can come talk to me during the Department of Psychological and Brain Sciences practice poster session. this poster session will be, scheduled for November 12 from 12:00 - 1:00 in 217 Ames.

Note: My homepage has moved to the LIMBS wiki.