Integrating Statistical and Machine Learning Approaches to Identify Receptive Field Structure in Neural Populations

Aug 5, 2026

Upload by Logic

Neural coding connects neuron or ensemble activity to how the brain processes information, and researchers often classify neurons by the variables they represent. Classical statistical model identification can be principled but may fail to scale efficiently for recordings from massive neural populations, while machine learning can analyze large datasets yet needs extensive training data and is harder to interpret. This dissertation proposes an integrated statistical-and-ML framework to infer neuronal coding properties, evaluated on rat hippocampus and PFC data from LFP and spiking recordings during a spatial alternation task.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages99

Reading Time4 h 9 min

AccessFree to download

Introduction

Modified Spline Regression

Population Coding in Rat Hippocampus and PFC

Machine Learning Model Identification Framework

Modified Spline Regression

Point Process-GLM Framework

Spline Functions

Boston University OpenBU [http://open. bu.edu](http://open. bu.edu) Theses & Dissertations Boston University Theses & Dissertations 2023 Integrating statistical and machine learning approaches to identify receptive field structure in neural populations [https://hdl.handle.net/2144/45484](https://hdl.handle.net/2144/45484)[ ](https://hdl.handle.net/2144/45484)Boston University BOSTON UNIVERSITY COLLEGE OF ENGINEERING Dissertation INTEGRATING STATISTICAL AND MACHINE LEARNING APPROACHES TO IDENTIFY RECEPTIVE FIELD STRUCTURE IN NEURAL POPULATIONS by MEHRAD SARMASHGHI B.S., Sharif University of Technology, 2015 Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy © 2023 by MEHRAD SARMASHGHI All rights reserved Approved by First Reader Uri T. Eden, PhD Professor of Mathematics and Statistics Second Reader Brian Kulis, PhD Associate Professor of Electrical and Computer Engineering Associate Professor of Systems Engineering Associate Professor of Computer Science Third Reader Ioannis Ch. Paschalidis, PhD Distinguished Professor of Engineering Professor of Electrical and Computer Engineering Professor of Systems Engineering Professor of Biomedical Engineering Founding Professor of Computing and Data Sciences Fourth Reader Mark A. Kramer, PhD Professor of Mathematics and Statistics Acknowledgments To my family who carved me to what I am today, to my dear friends for being therefor me no matter what, I owe it all to you. I would like to give my warmest thanks tomy advisor, Uri Eden. Working under your supervision was indeed a blessing. You have been very supportive and patient with me, I owe you a lot. I am grateful to my committee members, Prof. Ioannis Paschalidis, Prof. Brian Kulis and Prof. Mark Kramer who have provided me with their thoughtful comments. Your inputs definitely have improved this work. I am also very grateful to Elizabeth Flagg and Christine Ritzkowski for their assistance and heartwarming support, and Boston University for Funding. Finally, to everyone in the Division of Systems Engineering, Thanks for your support and encouragement! Mehrad Sarmashghi INTEGRATING STATISTICAL AND MACHINE LEARNING APPROACHES TO IDENTIFY RECEPTIVE FIELD STRUCTURE IN NEURAL POPULATIONS MEHRAD SARMASHGHI Boston University, College of Engineering, 2023 Major Professors: Uri T. Eden, PhD Professor of Mathematics and Statistics Brian Kulis, PhD Associate Professor of Electrical and Computer Engineering Associate Professor of Systems Engineering Associate Professor of Computer Science ABSTRACT Neural coding is essential for understanding how the activity of individual neurons or ensembles of neurons relates to cognitive processing of the world. Neurons can code for multiple variables simultaneously and neuroscientists are interested in classifying neurons based on the variables they represent. Building a model identification paradigm to identify neurons in terms of their coding properties is essential to understanding how the brain processes information. Statistical paradigms are capable of methodologically determining the factors influencing neural observations and assessing the quality of the resulting models to characterize and classify individual neurons. However, as neural recording technologies develop to produce data from massive populations, classical statistical methods often lack the computational efficiency required to handle such data. Machine learning (ML) approaches are known for enabling efficient large scale data analysis; however, they require huge training data sets, and model assessment and interpretation are more challenging than for classical statistical methods. To address these challenges, we develop an integrated framework, combining statistical modeling and machine learning approaches to identify the coding properties of neurons from large populations. In order to evaluate

FAQs

What problem does the dissertation address in neural population analysis?

It targets limitations of classical statistical methods when scaling to massive neural recordings, and the interpretability/training-data challenges of machine learning for identifying neuronal coding properties.

How is the proposed framework evaluated?

It is tested on data from rat hippocampus CA1 and prefrontal cortex during a spatial alternation task on a W-shaped track, using simultaneous local field potentials and spiking data.

What are the main components of the work across the related projects?

The dissertation improves spline-based statistical models, uses statistical model identification to classify neurons by coding properties, and trains a supervised CNN classifier using statistical-model outputs and additional simulated data.

Aug 5, 2026

Upload by Logic