Student Spotlight: Eszter Kish


kish 1

8 April 2015—Today’s Research Spotlight features Eszter Kish, a senior Neuroscience student in the College of Natural Sciences. Eszter is completing her Certificate in Scientific Computation with a research project supervised by Dr. Lizhen Lin, Assistant Professor in the Department of Statistics and Data Sciences.

The Certificate in Scientific Computation program is open to all undergraduate students interested in the use of mathematical, statistical, and computer-based techniques to investigate complex systems. The research project is the culminating requirement for the 18-credit program.

•   The project you are working on, the Certificate in Scientific Computation is titled, “Bayesian modeling of neuron firing rate maps using a B-spline basis prior.” What does the project entail?

The project entails implementing Bayesian inference of the neuron firing rate function with a B-spline basis prior. The non-negative constraint of the neuron firing rate function can easily be incorporated by imposing appropriate constraints on the coefficients of the spline basis.

We are using Gibbs sampling and a Metropolis-Hastings algorithm to sample from the posterior distribution, based on which statistical inference is carried out. We will implement the model using simulated data, in which one can compare the estimate with the ground truth, as well as a real data set. The final step will be to quantify uncertainties such as obtaining confidence intervals for the resulting estimates.

•    What is the goal of the project?​

The goal of the project is to build a flexible statistical model and efficient implementation schemes for estimating the neuron firing rates functions with single neuron data collected in many experimental studies of neuroscience.

•    What are the applications of your research beyond your specific project?

The statistical model proposed in our research has much broader applications beyond our project. In neuroscience, it is now routine to collect single neurons data under different levels of stimuli, experimental conditions, etc. Statistical inference of the population neuron firing function is crucial in understanding the neural functions under different conditions. Our proposed model can be applied in these broad settings.

•    How would you describe what you are doing to a lay audience?

Information in the brain can be encoded by the robustness of the firing of neurons. Neurons can have a preferred stimulus at which they fire the most. For example, in the visual system, some neurons have preferred colors that they respond to, while still increasing their firing to other colors, although not to the same extent. Therefore, the way neurons respond to stimuli can be described by a firing rate function into which we can put a stimulus and receive back an estimate of what a neuron’s actual response to that stimulus would look like.

•    What first interested you in this project?

I am interested in both statistics and neuroscience, and as this project combined the two disciplines I thought of it as a good way to learn more about both.

•   Why did you ask Dr. Lizhen Lin to supervise your project? What expertise does she bring?

Some of Dr. Lin’s fields of research are in Bayesian inference and machine learning, and she collaborated with a neuroscience lab while doing a postdoc at Duke. I wanted to learn more about Bayesian inference and statistical modeling, and I felt that Dr. Lin’s expertise in these areas, as well as her background in neuroscience, would help me explore these topics. 

•    Do you plan to continue this line of research in the future? How so?

Yes. Bayesian inference and MCMC sampling are foundations for statistical inference, and have broad applications in many fields of research. No matter what direction I end up going in the future, these statistical methods will be very useful.

Alumni Spotlight: Xiao Bai
SDS 328M Students Show Off Data Analysis Skills at...

Related Posts