TY - BOOK

T1 - Statistical Methods for Neural Data

T2 - Cointegration Analysis of Coupled Neurons & Generalized Linear Models for Spike Train Data

AU - Østergaard, Jacob

PY - 2017

Y1 - 2017

N2 - Some of the most captivating questions in the history of science concerns thefunctions of the human brain and the subject has attracted researchers and philosophersfor centuries. Recent advances in laboratory technology has enabled us tolook further into the internal microscopic components of the brain than ever before.Neuroscience, as a purely scientific discipline, is relatively new compared toit’s basic components of mathematics, physics, chemistry, and physiology. However,the current rate of experimental discoveries in neuroscience calls for new advancesin analytical tools to better understand the biological processes that occur inthe brain.This thesis aims to explore new statistical models for neural data and their usefulnessin analyzing experimental data. The thesis consist of two parts, one that concernsneural networks and how these can be interpreted as a cointegrated system andone that examines how the class of Generalized Linear Models can be used to decodespecific behaviors of simulated neurons. Part one introduces the concept of cointegrationand demonstrates how a network can be analyzed by interpreting the systemas a cointegrated process. This work is then extended from a small 3-dimensionalsystem to a high-dimensional setting and includes a discussion of future possibilitiesfor network analysis using these techniques. Part two opens with a demonstrationof how Generalized Linear Models can be designed for spike train data and howvarying patterns of different neurons are captured by this class of statistical models.Part two then continues with a specialized model aimed at capturing a specific typeof behavior known as "bursting".In the age of big data and artificial intelligence, two major themes related to neurosciencepresent themselves. The first is how to cope with the rapidly increasingdata collection from laboratory experiments and (very) high-dimensional interactingsystems. This occurs partially due to an increased interest in neuroscience, aswell as the introduction of new measuring equipment. The second is the motivationfor a continuously deeper understandning of the human brain. There are still countlessunanswered questions regarding this biological mechanism. In order to furtherunderstand causes of neural diseases as well as continued development of artificialintelligence, these questions are important to study. Ultimately, they should lead usto a better intuition regarding the question: "how does intelligence work"?

AB - Some of the most captivating questions in the history of science concerns thefunctions of the human brain and the subject has attracted researchers and philosophersfor centuries. Recent advances in laboratory technology has enabled us tolook further into the internal microscopic components of the brain than ever before.Neuroscience, as a purely scientific discipline, is relatively new compared toit’s basic components of mathematics, physics, chemistry, and physiology. However,the current rate of experimental discoveries in neuroscience calls for new advancesin analytical tools to better understand the biological processes that occur inthe brain.This thesis aims to explore new statistical models for neural data and their usefulnessin analyzing experimental data. The thesis consist of two parts, one that concernsneural networks and how these can be interpreted as a cointegrated system andone that examines how the class of Generalized Linear Models can be used to decodespecific behaviors of simulated neurons. Part one introduces the concept of cointegrationand demonstrates how a network can be analyzed by interpreting the systemas a cointegrated process. This work is then extended from a small 3-dimensionalsystem to a high-dimensional setting and includes a discussion of future possibilitiesfor network analysis using these techniques. Part two opens with a demonstrationof how Generalized Linear Models can be designed for spike train data and howvarying patterns of different neurons are captured by this class of statistical models.Part two then continues with a specialized model aimed at capturing a specific typeof behavior known as "bursting".In the age of big data and artificial intelligence, two major themes related to neurosciencepresent themselves. The first is how to cope with the rapidly increasingdata collection from laboratory experiments and (very) high-dimensional interactingsystems. This occurs partially due to an increased interest in neuroscience, aswell as the introduction of new measuring equipment. The second is the motivationfor a continuously deeper understandning of the human brain. There are still countlessunanswered questions regarding this biological mechanism. In order to furtherunderstand causes of neural diseases as well as continued development of artificialintelligence, these questions are important to study. Ultimately, they should lead usto a better intuition regarding the question: "how does intelligence work"?

UR - https://soeg.kb.dk/permalink/45KBDK_KGL/fbp0ps/alma99122123327305763

M3 - Ph.D. thesis

BT - Statistical Methods for Neural Data

PB - Department of Mathematical Sciences, Faculty of Science, University of Copenhagen

ER -