In this work, we develop models of neural circuits and plasticity rules that underlie different forms of learning and memory, with a focus on learning processes that involve multiple brain regions. We begin by surveying the literature on synaptic plasticity rules and implementations of learning algorithms in the brain. Each subsequent chapter presents a model of how a specific aspect of learning is implemented biologically, based on experimental evidence and normative considerations. We first focus on the neural basis of reinforcement learning in the basal ganglia. We show that in order to enable effective learning when control of behavior is distributed across multiple regions (``off-policy reinforcement learning''), classic models of dopamine activity must be adapted to include an additional action-sensitive component. We also show that the known plasticity rules of direct and indirect-pathway striatal projection neurons are inconsistent with existing models of striatal codes for action. We propose and find experimental support for a new model of striatal activity driven by efferent input. This model is functionally compatible with striatal plasticity rules and enables simultaneous multiplexing of action-selection and learning signals, a necessary ingredient for off-policy reinforcement learning. We next use an off-policy reinforcement learning model to explain a new experimental finding about the conditions under which learned motor skills are consolidated to be driven by the dorsolateral striatum in rats. We then shift our focus to consider consolidation more broadly, proposing a general model of the advantages of systems in which memories and learned behaviors are consolidated from short-term to long-term learning pathways. In particular, our model proposes that such architectures enable selective filtering of the set of experiences used for learning, which can be essential in noisy environments with many extraneous stimuli. In the appendices, we explore other factors relevant to learning algorithms, including the interaction between multiple sensory modalities, and the problem of credit assignment in multi-layer neural networks. In summary, this work presents a varied set of models of different forms of learning in the brain, emphasizing the cooperative role of plasticity rules and multi-regional circuit architecture in producing functionally useful synaptic weight updates.

Play fighting is common in juvenile mammals as a peri-pubertal form of agonistic behavior preceding adult aggressive behavior. In golden hamsters, play fighting peaks in early puberty around postnatal day 35 (P-35), and gradually matures into adult aggression in late puberty. Though extensively studied, the neural mechanisms underlying play fighting remains poorly understood. My dissertation focuses on identifying the neural circuitry and neural transmitter systems that mediate this behavior in juvenile golden hamsters. Based on behavioral similarities between the offensive components of play fighting and adult aggression, I predicted that the neural circuitries mediating both behaviors shared common components. This possibility was tested by quantifying the immunolabeling of c-Fos expression in juvenile hamsters after the consummation of play fighting. In support of my hypothesis, I found that areas previously associated with offensive aggression in adult hamsters, including the ventrolateral hypothalamus (VLH), the posterior dorsal part of the medial amygdala (MePD), and the bed nucleus of the stria terminalis (BST), also showed enhanced c-Fos expression after play fighting, which supported my hypothesis. Vasopressin (AVP) facilitates aggression in adult hamsters. Therefore, I hypothesized that AVP also activates play fighting. To test my hypothesis, juvenile male golden hamsters were tested for play fighting after they received central microinjections of an AVP V1A-receptor antagonist into the anterior hypothalamus (AH). Also, immunocytochemistry was performed to identify possible AVP neurons associated with this behavior. I found that the AVP antagonist selectively inhibited the attack components of play fighting in experimental animals. In addition, AVP cells in the nucleus circularis (NC) and the medial division of the supraoptic nucleus (mSON), which were associated with offensive aggression, also showed increased c-Fos activity after play fighting. Together, these results show that AVP facilitates offensive behaviors throughout hamster development, from play fighting in juveniles to aggression in adults. A recent study shows that oral administration of a CRH receptor antagonist inhibits aggression in adult hamsters. Therefore, I predicted that CRH plays a similar role in play fighting. To test my prediction, juvenile hamsters were tested for play fighting after central microinjections of a CRH receptor antagonist. I found that microinjections of the CRH receptor antagonist within the lateral septum (LS) resulted in an inhibition of several aspects of play fighting. The possible source of CRH affecting the behavior was tested through combined immunocytochemistry to CRH and c-Fos. I found CRH neurons in the diagonal band of Broca (DBB), an area with extensive connections with the LS, were particularly activated in association with play fighting. In conclusion, I find that shared neural elements participating in the “vertebrate social behavior neural network” are associated with both aggression and play fighting in hamsters. This circuitry is activated before the onset of puberty and is affected by rising levels of steroid hormones during the developmental period leading to adult behaviors. Within the circuitry, vasopressin release in the AH appears to control the activation of play fighting attacks. In contrast, CRH release in the LS affects a broader range of aspects of play fighting, including not just consummatory aspects of the behavior, but apparently also appetitive components in the form of contact duration. / text

Relações entre machos de roedores muitas vezes se dão com a manifestação de comportamentos agressivos, em um embate em que se define um perdedor e um ganhador. O macho perdedor, de acordo com experimentos realizados no laboratório, apresenta comportamentos de defesa (avaliação de risco) quando exposto ao contexto da derrota social. Na expressão da defesa condicionada, estruturas do hipotálamo lateral devem ser importantes, que são também mobilizadas durante o confronto. No hipotálamo lateral destacamos a a parte justadorsomedial do hipotálamo lateral (LHAjd) que tem conexões com o sistema septo-hipocampal e projeta-se liga ao prémamilar dorsal (PMD), crítico para a expressão de comportamentos de defesa. Por outro lado, o PMD, que é uma estrutura altamente mobilizada durante o confronto social, tem conexões com o núcleo anteromedial do tálamo (AMv), estrutura já conhecida pela sua importância na aquisição da memória contextual e espacial. Assim, no presente trabalho, investigamos o papel do LHAjd na expressão e do AMv na aquisição da defesa condicionada na derrota social. Nos animais com lesão do LHAjd, observamos diminuição nos comportamentos de avaliação de risco durante exposição ao contexto. Nos animais com lesão do AMv o mesmo efeito. Dado os efeitos das lesões no AMv, elaboramos um paradigma para estudo em camundongos, a fim de se realizar inativações pontuais com farmacogenética nos neurônios glutamatérgicos em animais transgênicos durante a derrota social. Foi observado o mesmo padrão em camundongos transgênicos vGlut2-cre inativados com farmacogenética, não havendo influência nos comportamentos durante a derrota social. Inativações antes do contexto não causaram efeito na defesa condicionada. As inativações durante a derrota, no entanto, não tiveram efeito quando o residente agressivo está na exposição ao contexto. Os dados sugerem que o LHAjd tem papel na expressão da defesa condicionada, enquanto que o AMv tem papel na aquisição da defesa condicionada ao contexto, mas não no reconhecimento do residente agressivo. / Male rodents may interact aggressively, and from the agonistic encounter, it results a winner and a defeated animal. Accordingingly, the defeated male shows defensive behaviors (risk assessment) to the social defeat-related context. Contextual responses are known to rely on hippocampal processing, and one of the main targets of the hippocampal system is the justodorsomedial part of lateral hypothalamus (LHAjd), which projects to dorsal premamillary nucleus (PMD), known to be involved in the expression of social defensive behaviors. Notably, PMD, a hypothalamic site highly responsive to the social defeat, in turn, projects to the ventral part of thalamic anteromedial nucleus (AMv), previously shown to be involved in the acquistion of spatial and contextual memory to predatory threats. Thus, in present study, we investigated the role of LHAjd in expression and of AMv in acquisition of social defeat conditioned defensive behaviors. In LHAjd lesioned animals, we observed a decrease in risk assessment behaviors during exposute to the social defeat associated context , but not during the social defeat itself, suggesting a role in the expression of contextual but not in the innate social defeat. In AMv lesioned animals, we observed that the animals lost contextual defensive response, suggesting a role in the acquisition and/or expression of contextual responses. Next, using pharmacogetic inhibition, we investigated in vGlut2-cre transgenic mice the role of the AMV in the acquisition and expression of contextual defensive behavior. We have found that AMV inactivation prior to the social defeaf, but not prior to the exposure to the social defeat related context, was able to decrease contextual responses in animals tested withot the presence of the male aggressor, but not in the situation where the male aggressor was present. Overall, our results suggest that the LHAjd has a role in the expression of conditioned defense, and that the AMV is involved in the acquisision of contextual fear responses, but not in social recognition of aggressive male.

The work presented in this thesis provides new perspectives and approaches for problems that arise in the analysis of neural data. Particular emphasis is placed on parameter fitting and automated analysis problems that would arise naturally in closed-loop experiments. Part one focuses on two brain-computer interface problems. First, we provide a framework for understanding co-adaptation, the setting in which decoder updating and user learning occur simultaneously. We also provide a new perspective on intention-based parameter fitting and tools to extend this approach to higher dimensional decoders. Part two focuses on event inference, which refers to the decomposition of observed timeseries data into interpretable events. We present application of event inference methods on voltage-clamp recordings as well as calcium imaging, and describe extensions to allow for combining data across modalities or trials.

Ho, Tsz Wan. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2009. / Includes bibliographical references (leaves 125-165). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese.

This research presents a new performance improvement technique, window memoization, for software and hardware implementations of local image processing algorithms. Window memoization combines the memoization techniques proposed in software and hardware with a characteristic of image data, computational redundancy, to improve the performance (in software) and efficiency (in hardware) of local image processing algorithms. The computational redundancy of an image indicates the percentage of computations that can be skipped when performing a local image processing algorithm on the image. Our studies show that computational redundancy is inherited from two principal redundancies in image data: coding redundancy and interpixel redundancy. We have shown mathematically that the amount of coding and interpixel redundancy of an image has a positive effect on the computational redundancy of the image where a higher coding and interpixel redundancy leads to a higher computational redundancy. We have also demonstrated (mathematically and empirically) that the amount of coding and interpixel redundancy of an image has a positive effect on the speedup obtained for the image by window memoization in both software and hardware. Window memoization minimizes the number of redundant computations performed on an image by identifying similar neighborhoods of pixels in the image. It uses a memory, reuse table, to store the results of previously performed computations. When a set of computations has to be performed for the first time, the computations are performed and the corresponding result is stored in the reuse table. When the same set of computations has to be performed again in the future, the previously calculated result is reused and the actual computations are skipped. Implementing the window memoization technique in software speeds up the computations required to complete an image processing task. In software, we have developed an optimized architecture for window memoization and applied it to six image processing algorithms: Canny edge detector, morphological gradient, Kirsch edge detector, Trajkovic corner detector, median filter, and local variance. The typical speedups range from 1.2 to 7.9 with a maximum factor of 40. We have also presented a performance model to predict the speedups obtained by window memoization in software. In hardware, we have developed an optimized architecture that embodies the window memoization technique. Our hardware design for window memoization achieves high speedups with an overhead in hardware area that is significantly less than that of conventional performance improvement techniques. As case studies in hardware, we have applied window memoization to the Kirsch edge detector and median filter. The typical and maximum speedup factors in hardware are 1.6 and 1.8, respectively, with 40% less hardware in comparison to conventional optimization techniques.

This research presents a new performance improvement technique, window memoization, for software and hardware implementations of local image processing algorithms. Window memoization combines the memoization techniques proposed in software and hardware with a characteristic of image data, computational redundancy, to improve the performance (in software) and efficiency (in hardware) of local image processing algorithms. The computational redundancy of an image indicates the percentage of computations that can be skipped when performing a local image processing algorithm on the image. Our studies show that computational redundancy is inherited from two principal redundancies in image data: coding redundancy and interpixel redundancy. We have shown mathematically that the amount of coding and interpixel redundancy of an image has a positive effect on the computational redundancy of the image where a higher coding and interpixel redundancy leads to a higher computational redundancy. We have also demonstrated (mathematically and empirically) that the amount of coding and interpixel redundancy of an image has a positive effect on the speedup obtained for the image by window memoization in both software and hardware. Window memoization minimizes the number of redundant computations performed on an image by identifying similar neighborhoods of pixels in the image. It uses a memory, reuse table, to store the results of previously performed computations. When a set of computations has to be performed for the first time, the computations are performed and the corresponding result is stored in the reuse table. When the same set of computations has to be performed again in the future, the previously calculated result is reused and the actual computations are skipped. Implementing the window memoization technique in software speeds up the computations required to complete an image processing task. In software, we have developed an optimized architecture for window memoization and applied it to six image processing algorithms: Canny edge detector, morphological gradient, Kirsch edge detector, Trajkovic corner detector, median filter, and local variance. The typical speedups range from 1.2 to 7.9 with a maximum factor of 40. We have also presented a performance model to predict the speedups obtained by window memoization in software. In hardware, we have developed an optimized architecture that embodies the window memoization technique. Our hardware design for window memoization achieves high speedups with an overhead in hardware area that is significantly less than that of conventional performance improvement techniques. As case studies in hardware, we have applied window memoization to the Kirsch edge detector and median filter. The typical and maximum speedup factors in hardware are 1.6 and 1.8, respectively, with 40% less hardware in comparison to conventional optimization techniques.

The extent to which cortical maps may reorganize in adult humans is a significant and topical debate in visual neuroscience. Though there are conflicting findings, evidence from humans and animals indicates that the topography of the visual cortex may change after retinal deafferentation. Remarkably, this reorganization seems to be possible in adults, whose brains are less amenable to plastic change. If adult visual reorganization is legitimate, an understanding of its causes and consequences could be profound considering the millions suffering from age-related visual disorders. This dissertation explores whether visual training may yield a reorganization of sensory maps in the adult visual cortex. It describes research in which patients, diagnosed with age-related macular degeneration (AMD), underwent visual rehabilitation therapy. Functional brain scans and behavioral tests were conducted pre and post training. These interventions generated valuable knowledge regarding whether "reorganized" activity is a true rewiring of feed forward cortical processes or an artifact of attentional feedback. The rehabilitation training produced demonstrable differences in activation patterns along the primary visual cortex (V1), but sparse improvement in the behavioral tests. In contrast, there was significant improvement in fixation tests which assessed oculomotor control. These results suggest that the nature of reorganized activity has more to do with attentional mechanisms than feed forward reorganization. Future investigations could benefit from examining the brain sites that govern visual attention in the frontal and parietal cortices. These areas may have more to do with visual adaptation in AMD patients than V1.

Activity-induced immediate-early gene (IEG) transcription foci can be labelled with fluorescent probes, permitting high temporal and spatial resolution in mapping neuronal circuits. Previous quantification approaches have assumed a Boolean function of transcription foci, assuming that cells are either active or inactive. Due to multiple amplification steps in the in situ hybridization process, it was thought that information relating to magnitudes of firing rates was lost. However, the current data suggest that transcription foci actually exhibit non-Boolean intensity and size values which vary according to behavioural condition. Systematic characterization of transcription foci intensity and size revealed incremental variations such that: home-cage < one-environment exposure < five-environment exposure < maximal electroconvulsive shock. Visual differences in transcription foci may result from a quantifiable relationship between spiking patterns and transcription rates. The exact stoichiometry between neuronal spiking and transcription is not yet clear, but these results suggest that Boolean applications of IEG imaging may neglect accurate neuronal activation properties. / xvi, 125 leaves : ill. ; 29 cm

This dissertation research explored the capabilities of neuronal networks grown on substrate integrated microelectrode arrays in vitro to be applied to toxicological research and lipid signaling. Chapter 1 details the effects of chlorine on neuronal network spontaneous electrical activity and pharmacological sensitivity. This study demonstrates that neuronal networks can maintain baseline spontaneous activity, and respond normally to pharmacological manipulations in the present of three times the chlorine present in drinking water. The findings suggest that neuronal networks may be used as biological sensors to monitor the quality of water and the presence of novel toxicants that cannot be detected by conventional sensors. Chapter 2 details the neuromodulatory effects of N-acylethanolamides (NAEs) on the spontaneous electrical activity of neuronal networks. NAEs are a group of lipids that can mimic the effects of marijuana and can be derived from a variety of plant sources including soy lecithin. The most prominent NAEs in soy lecithin, palmitoylethanolamide (PEA) and linoleoylethanolamide (LEA), were tested individually and were found to significantly inhibit neuronal spiking and bursting activity. These effects were potentiated by a mixture of NAEs as found in a HPLC enriched fraction from soy lecithin. Cannabinoid receptor-1 (CB1-R) antagonists and other cannabinoid pathway modulators indicated that the CB1-R was not directly involved in the effects of NAEs, but that enzymatic degradation and cellular uptake were more likely targets. The results demonstrate that neuronal networks may also be a viable platform for the elucidation of biochemical pathways and drug mechanisms of action.