@phdthesis{10.7907/hmyr-3d58, author = {Rosenberg, Matthew Hutson}, title = {Innate Navigation: Magnetic Sensation and Maze Learning}, school = {California Institute of Technology}, year = {2022}, doi = {10.7907/hmyr-3d58}, url = {https://resolver.caltech.edu/CaltechTHESIS:06072022-184731596}, abstract = {
This thesis aims to advance the understanding of the neurobiology of navigation through the investigation of two topics: magnetic sensation and maze navigation. The central question of this work may be framed as follows: how do animals find their way to key resources that are necessary for survival? Three projects are presented to address this.
Chapter II explores a sensory hypothesis that some animals may navigate long distances by directly sensing the earth’s magnetic field. Awake zebra finches were stimulated with magnetic fields that varied sinusoidally in time while electrical recordings were collected via multi-channel electrodes. Preliminary negative results are presented, along with a detailed statistical treatment indicating no significant effect of magnetic stimulation on neural activity.
Chapter III presents a novel approach to studying learning and navigation in animal subjects. Mice are allowed free passage between a normal home cage and a complex maze environment, coming and going as they please. Sated animals, with free access to food and water, spend significant portions of a given multi-hour experiment in the maze and display efficient exploration. Water-restricted animals show three additional phenomena: immediate knowledge of the route home, rapid learning of the location of a single water port among 64 similar locations, and a moment of “sudden insight” in which the rate at which long, direct routes to the water source, beginning from many locations, increases discontinuously.
Chapter IV offers a simple, biologically feasible circuit model that recapitulates and explains some of the rapid learning behaviors we observe in mice. This model suggests a mechanism that might allow mice to flexibly store and recall direct routes to different resources that are activated by different internal drives.
The final chapter outlines some potential directions for future inquiry, including potential maze experiments to conduct with wireless electrophysiology and expansion of the range of species tested for magnetic perception. The Appendix briefly describes some follow-up experiments and intriguing preliminary results. Similarities in the navigation deficit displayed by mice that have been experimentally perturbed in several disparate ways is noted briefly. These perturbations include whisker trimming, olfactory neuron ablation, genetic ablation of cortex and hippocampus, and opiate intoxication.
}, address = {1200 East California Boulevard, Pasadena, California 91125}, advisor = {Meister, Markus}, } @phdthesis{10.7907/rj2p-8g11, author = {Bagherian, Dawna Paria}, title = {Artificial Neural Networks for Nonlinear System Identification of Neuronal Microcircuits}, school = {California Institute of Technology}, year = {2021}, doi = {10.7907/rj2p-8g11}, url = {https://resolver.caltech.edu/CaltechTHESIS:05282021-174607976}, abstract = {This thesis explores the application of artificial neural networks (ANNs) to nonlinear system identification. We use neuronal microcircuits in the retina as a testbed for our technique, which relies upon the marriage of partial anatomical information with large electrophysiological datasets. Rather than a typical application of machine learning, our primary goal is not to predict the output of retinal circuits, but rather to uncover their structure. We begin with a theoretical exploration in a toy problem and provide a proof of unique identifiability under a specific set of conditions. We then perform empirical simulations in a number of different circuit architectures and explore the space of constraints and regularizers to demonstrate that this technique is feasible in a hyperparametric regime that lends itself well to neuroscience datasets. We then apply the technique to mouse retinal datasets and show that we can both recover known biological information as well as discover new hypotheses for biological exploration. We end with an exploration of active stimulus design algorithms to distinguish between circuit hypotheses.
}, address = {1200 East California Boulevard, Pasadena, California 91125}, } @phdthesis{10.7907/czd1-dp02, author = {Turan, Zeynep}, title = {Life Without Cortex: Subcortical Circuits in Naturalistic Behaviors}, school = {California Institute of Technology}, year = {2021}, doi = {10.7907/czd1-dp02}, url = {https://resolver.caltech.edu/CaltechTHESIS:06082021-032229529}, abstract = {A major goal of neuroscience is to understand the neural circuits underlying animal behavior. Many contemporary studies focus on behavioral tasks which do not reflect realistic conditions, such as mapping an arbitrary sensory stimulus to motor output. Given that the brain evolved within the context of the natural environment, it is more likely that these circuits were optimized for naturalistic behaviors such as avoiding predators, hunting, and social interactions with conspecifics. Many of these naturalistic behaviors predate the great expansion of the neocortex in mammals, as they are crucial for the survival of any animal. Using a mutant mouse model and surgical techniques, we show that the evolutionarily ancient subcortical circuits of mice are sufficient for sensory processing, stimulus discrimination, and exhibiting robust innate defensive behaviors in a predator avoidance assay. Furthermore, these animals are capable of navigating a complex labyrinth, which challenges long-held beliefs that learning and memory require the neocortex and the hippocampus. Our results emphasize the significant capacity of subcortical circuits in behaviors necessary for survival and illustrate the importance of using naturalistic behaviors to probe brain function.}, address = {1200 East California Boulevard, Pasadena, California 91125}, advisor = {Meister, Markus}, } @phdthesis{10.7907/sq58-z682, author = {Liu, Yang}, title = {From Restoring Human Vision to Enhancing Computer Vision}, school = {California Institute of Technology}, year = {2020}, doi = {10.7907/sq58-z682}, url = {https://resolver.caltech.edu/CaltechTHESIS:06092020-120629159}, abstract = {The central theme of this work is enabling vision, which includes two subtopics: restoring vision for blind humans, and enhancing computer vision models in visual recognition. Chapter 1 first provides a gentle introduction to relevant high level principles of human visual computations and summarizes two fundamental questions that vision answers: “what” and “where.” Chapters 2, 3, and 4 contain three published projects that are anchored by those two fundamental questions.
Chapter 2 introduces a cognitive assistant to restore visual function for blind humans by focusing on an interface powered by audio augmented reality. The assistant communicates the “what” and “where” aspects of visual scenes by a combination of natural language and spatialized sound. We experimentally demonstrated that the assistant enables many aspects of visual functions for naive blind users.
Chapters 3 and 4 develop data augmentation methods to address the data inefficiency problem in neural network based computer visual recognition models. In Chapter 3, a 3D-simulation based data augmentation method is developed for improving the generalization of visual classification models for rare classes. In Chapter 4, a fast and efficient data augmentation method is developed for the newly formulated panoptic segmentation task. The method improves performance of state-of-the-art panoptic segmentation models and generalizes across dataset domains, sizes, model architectures, and backbones.
}, address = {1200 East California Boulevard, Pasadena, California 91125}, advisor = {Meister, Markus}, } @phdthesis{10.7907/zn2j-m319, author = {Lee, Kyu Hyun}, title = {Visual Computations in the Superior Colliculus}, school = {California Institute of Technology}, year = {2020}, doi = {10.7907/zn2j-m319}, url = {https://resolver.caltech.edu/CaltechTHESIS:06012020-165101402}, abstract = {This thesis presents two projects related to large-scale extracellular recordings of neural signals. The first project asks how the brain sifts the onslaught of sensory information to identify the few bits that are relevant for guiding behavior. This question is studied in the context of the looming reaction, an innate defensive behavior against an approaching aerial predator. Interestingly, the mouse responds very selectively to the looming stimulus regardless of changes in orthogonal features, such as its position. The neural basis of this phenomenon is investigated with extracellular recordings in the superior colliculus, a midbrain visual area known to mediate the looming reaction. A detailed analysis of the difference between the superficial and deeper layers of the superior colliculus highlights a core function of visual processing: to discard information intelligently.
The second project presents electrode pooling, a novel method to increase the yield of extracellular recordings with the modern silicon electrode array. The fundamental constraint of wire volume in these devices is identified, and a solution that makes use of the switching circuitry and the sparseness of the neural signal in the time axis is described. Specifically, the method proposes to intelligently choose many recording sites that carry signal and connect them to a single wire via manipulating the switches. This pooled recording is subsequently un-mixed by a spike-sorting algorithm. The method is implemented in a state-of-the-art silicon neural probe, and its effect on signal and noise is analyzed by theory and experiment. Recommendations on the design of silicon devices are made to facilitate the incorporation of this method in the future.
}, address = {1200 East California Boulevard, Pasadena, California 91125}, advisor = {Meister, Markus}, } @phdthesis{10.7907/Z92R3PMW, author = {Shai, Adam S.}, title = {The Physiology and Computation of Pyramidal Neurons}, school = {California Institute of Technology}, year = {2016}, doi = {10.7907/Z92R3PMW}, url = {https://resolver.caltech.edu/CaltechTHESIS:01042016-124746578}, abstract = {A variety of neural signals have been measured as correlates to consciousness. In particular, late current sinks in layer 1, distributed activity across the cortex, and feedback processing have all been implicated. What are the physiological underpinnings of these signals? What computational role do they play in the brain? Why do they correlate to consciousness? This thesis begins to answer these questions by focusing on the pyramidal neuron. As the primary communicator of long-range feedforward and feedback signals in the cortex, the pyramidal neuron is set up to play an important role in establishing distributed representations. Additionally, the dendritic extent, reaching layer 1, is well situated to receive feedback inputs and contribute to current sinks in the upper layers. An investigation of pyramidal neuron physiology is therefore necessary to understand how the brain creates, and potentially uses, the neural correlates of consciousness. An important part of this thesis will be in establishing the computational role that dendritic physiology plays. In order to do this, a combined experimental and modeling approach is used.
This thesis beings with single-cell experiments in layer 5 and layer 2/3 pyramidal neurons. In both cases, dendritic nonlinearities are characterized and found to be integral regulators of neural output. Particular attention is paid to calcium spikes and NMDA spikes, which both exist in the apical dendrites, considerable distances from the spike initiation zone. These experiments are then used to create detailed multicompartmental models. These models are used to test hypothesis regarding spatial distribution of membrane channels, to quantify the effects of certain experimental manipulations, and to establish the computational properties of the single cell. We find that the pyramidal neuron physiology can carry out a coincidence detection mechanism. Further abstraction of these models reveals potential mechanisms for spike time control, frequency modulation, and tuning. Finally, a set of experiments are carried out to establish the effect of long-range feedback inputs onto the pyramidal neuron. A final discussion then explores a potential way in which the physiology of pyramidal neurons can establish distributed representations, and contribute to consciousness.
}, address = {1200 East California Boulevard, Pasadena, California 91125}, advisor = {Meister, Markus}, } @phdthesis{10.7907/PANC-K035, author = {Mazor, Ofer}, title = {Neural Dynamics and Population Coding in the Insect Brain}, school = {California Institute of Technology}, year = {2005}, doi = {10.7907/PANC-K035}, url = {https://resolver.caltech.edu/CaltechETD:etd-06062005-113150}, abstract = {Sensory information is represented in the brain through the activity of populations of neurons. How this information is encoded and how it is processed and read out are crucial questions in neuroscience. The work presented here examines these issues using an insect brain model system. Specifically, this work addresses how odor information is represented across a population of neurons in this relatively simple nervous system. It asks how the dynamics of a population of neurons contribute to the encoding of information.
To address these questions, simultaneous multi-unit extracellular recordings were made in vivo in the locust brain. The first part of the dissertation describes several advances in spike-sorting methods that were necessary for analyzing such recordings. These advances include quantitative tests of sorting quality, and they allow for automated spike-sorting. Using these techniques, data sampled from tens of neurons over hours of recording can be analyzed with relative ease.
The remainder of the dissertation examines the encoding of olfactory information by a population of neurons called projection neurons (PNs), located in the first olfactory relay of the brain. Odor information is shown to be represented by a subpopulation of responsive PNs. The composition of this population changes over time in an odor-specific manner, thus forming a distributed, dynamical representation. The statistics of this response and its dynamics are quantified.
Furthermore, the mechanism by which odor information is extracted from the PN population response is examined. A second set of recordings were made from Kenyon cells (KCs), which receive direct excitatory synaptic input from PNs. The dynamic response of the PN population appears to be decoded by KCs through a mechanism based on several underlying components, including oscillatory dynamics, feed-forward inhibition, and intrinsic properties of the KCs. This decoding process is shown to drastically change the odor representations, from dense to sparse.
Taken together, the results presented in this dissertation establish that the complex spatial and temporal dynamics of the PN population do encode odor information, and that this information is decoded by other neurons (KCs) in a very precise way, resulting in a drastic transformation of representation. The basic mechanisms underlying this transformation exist in many brain areas and across phyla, suggesting that many of the principles described here could be of general relevance.
}, address = {1200 East California Boulevard, Pasadena, California 91125}, advisor = {Laurent, Gilles J.}, }