Thesis (M.S.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1989. / Includes bibliographical references (leaves 112-116). / by Michael Drumheller. / M.S.

Thesis (M.S.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1991. / Includes bibliographical references (leaves 57-60). / by Anselm Spoerri. / M.S.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1992. / Includes bibliographical references. / by Stefan Treue. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Brain and Cognitive Sciences, 2018. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 131-143). / This thesis develops new approaches for three problems in machine learning, using tools from the study of optimal transport (or Wasserstein) distances between probability distributions. Optimal transport distances capture an intuitive notion of similarity between distributions, by incorporating the underlying geometry of the domain of the distributions. Despite their intuitive appeal, optimal transport distances are often difficult to apply in practice, as computing them requires solving a costly optimization problem. In each setting studied here, we describe a numerical method that overcomes this computational bottleneck and enables scaling to real data. In the first part, we consider the problem of multi-output learning in the presence of a metric on the output domain. We develop a loss function that measures the Wasserstein distance between the prediction and ground truth, and describe an efficient learning algorithm based on entropic regularization of the optimal transport problem. We additionally propose a novel extension of the Wasserstein distance from probability measures to unnormalized measures, which is applicable in settings where the ground truth is not naturally expressed as a probability distribution. We show statistical learning bounds for both the Wasserstein loss and its unnormalized counterpart. The Wasserstein loss can encourage smoothness of the predictions with respect to a chosen metric on the output space. We demonstrate this property on a real-data image tagging problem, outperforming a baseline that doesn't use the metric. In the second part, we consider the probabilistic inference problem for diffusion processes. Such processes model a variety of stochastic phenomena and appear often in continuous-time state space models. Exact inference for diffusion processes is generally intractable. In this work, we describe a novel approximate inference method, which is based on a characterization of the diffusion as following a gradient flow in a space of probability densities endowed with a Wasserstein metric. Existing methods for computing this Wasserstein gradient flow rely on discretizing the underlying domain of the diffusion, prohibiting their application to problems in more than several dimensions. In the current work, we propose a novel algorithm for computing a Wasserstein gradient flow that operates directly in a space of continuous functions, free of any underlying mesh. We apply our approximate gradient flow to the problem of filtering a diffusion, showing superior performance where standard filters struggle. Finally, we study the ecological inference problem, which is that of reasoning from aggregate measurements of a population to inferences about the individual behaviors of its members. This problem arises often when dealing with data from economics and political sciences, such as when attempting to infer the demographic breakdown of votes for each political party, given only the aggregate demographic and vote counts separately. Ecological inference is generally ill-posed, and requires prior information to distinguish a unique solution. We propose a novel, general framework for ecological inference that allows for a variety of priors and enables efficient computation of the most probable solution. Unlike previous methods, which rely on Monte Carlo estimates of the posterior, our inference procedure uses an efficient fixed point iteration that is linearly convergent. Given suitable prior information, our method can achieve more accurate inferences than existing methods. We additionally explore a sampling algorithm for estimating credible regions. / by Charles Frogner. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1989. / Includes bibliographical references (leaves 151-158). / by Richard Patrick Wildes. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1997. / Includes bibliographical references (p. 335-338). / by John Francis Houde. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 2009. / Cataloged from PDF version of thesis. / Includes bibliographical references. / The basal ganglia form the largest sub-cortical structure in the human brain and are implicated in numerous human diseases. In songbirds, as in mammals, basal ganglia-forebrain circuits are necessary for the learning and production of complex motor behaviors; however, the precise role of this circuitry remains unknown. This thesis is an investigation into how the anterior forebrain pathway (AFP), an avian basal ganglia-forebrain circuit, supports vocal learning in the songbird. This investigation reveals two previously undiscovered functions of the AFP - both related to reinforcement, or trial-and-error, learning. One necessary component of reinforcement learning is the generation of variable behavior. The songs of learning juvenile birds are naturally highly variable. Rapid pharmacological inactivation of the AFP output nucleus causes an immediate and dramatic reduction in this variability. In addition, the first single-unit recordings of AFP output neurons in singing juvenile birds reveal little correlation with plastic song and a premotor correlation with the most variable form of singing, subsong. These results suggest a novel function for basal ganglia-forebrain circuitry in the production of exploratory behavior. A second component of reinforcement learning is the evaluation of performance-based feedback - auditory feedback in the case of singing. Conditional disruptive auditory feedback is a novel behavioral paradigm capable of causing rapid experimentally-controlled vocal learning. Inactivating the AFP while using this new paradigm to induce learning reveals that the AFP biases motor output to improve auditory feedback. This result suggests that basal gangliaforebrain circuits are involved in the evaluation of performance-based feedback. It also suggests for the first time that these circuits are capable of producing temporally precise premotor drive that incrementally improves a motor skill. In summary, this investigation significantly furthers the view that basal ganglia-forebrain circuitry is involved in reinforcement learning. It ascribes two functions to the anterior forebrain pathway: to drive variable behavior; and to bias future behavior incrementally towards better performance. By analogy, basal ganglia-thalamocortical loops may perform similar functions in mammals. / by Aaron Samuel Andalman. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Brain and Cognitive Sciences, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 158-177). / Huntington's disease (HD), the most common inherited neurodegenerative disorder, is caused by mutations in the huntingtin (HTT) gene, which encodes a poly-glutamine (polyQ) repeat protein. Despite widespread expression of the HTT gene, HD presents with massive neuronal cell loss and transcriptional dysregulation primarily in the striatum and deep layers of the cortex. Synaptic dysfunction and motor deficits are also prominent in HD patients as well as mouse models. In an attempt to identify factors that could both explain these alterations and mirror these vulnerability patterns, we identified a potential role for the striatal-enriched polyQ protein Foxp2 in HD. The transcription factor Foxp2 was recently identified as a crucial regulator of striatal synaptogenesis and corticostriatal inputs during striatal development, and has also been demonstrated to play a critical role in motor learning. Here we show that, in mice, overexpression of Foxp2 in the adult striatum of two models of HD leads to rescue of HD-associated behaviors, while knockdown of Foxp2 in wild-type adult striatum leads to development of HD-associated behaviors. We note that Foxp2 encodes the longest polyglutamine repeat protein in the human reference genome, and we show that it can be sequestered into aggregates with polyglutamine-expanded mutant HTT protein. Foxp2 overexpression in HD model mice leads to altered expression of several genes associated with synaptic function, genes which present new targets for normalization of corticostriatal dysfunction in HD. / by Lea June Hachigian. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1995. / Includes bibliographical references (p. 163-170). / by Ann Senghas. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1996. / Includes bibliographical references (p. 249-252). / by Kevin J. Broihier. / Ph.D.