Please join us for a series of short talks given by our BCS Graduate Students and new BCS Faculty Members

4:00 - 5:00PM Graduate Students

Presenter: Lou Beaulieu-Laroche, Harnett Lab

Talk Title: Biophysical mechanisms of dendritic integration in human neurons

Abstract: The biophysical features of neurons shape information processing in the brain. Cortical dendrites are much longer in humans than in other species, but it is unclear how their size affects synaptic integration. Here, we perform the first direct electrical recordings from human dendrites and report enhanced electrical compartmentalization. We find that the increased length of human dendrites alters their input-output properties, which will impact cortical computation.

Presenter: Amanda Vernon, Heiman Lab

Talk Title: Enhanced Striatal Glutamatergic Function Upon Chronic Antipsychotic Action

Abstract: Schizophrenia is a debilitating psychiatric disorder affecting 1% of the global population. Currently available antipsychotic drugs (APD) mitigate some symptoms in a subset of patients, but they frequently lead to significant side effects, and there is no preventive treatment or cure. Efforts to develop novel therapeutics are constrained by the fact that the therapeutic effect of currently prescribed APD is not well understood, and the cell type(s) crucial for beneficial effects have not been conclusively identified.  This study aims to improve our understanding of the molecular and cellular effects of APD to provide a path for improved therapy design and insight into the pathophysiology of the disorder itself.  We find that chronic APD administration leads to translational alterations that predict glutamatergic function enhancement in striatal spiny projection neurons. In line with this result, we also find that chronic APD administration leads to new synapses with calcium-permeable AMPA receptors in the ventral striatum. Finally, we observe that striatal astrocytes express comparatively higher levels of dopamine receptors and downregulate markers of reactivity in response to APD administration. This study reveals a core molecular signature of APD administration and suggests that enhancement of glutamatergic function in the ventral striatum may be an improved APD development strategy.

Presenter: Meilin Zhan, Levy Lab

Talk Title: Investigating Theories of Speaker Choice in a Classifier Language

Abstract: The simple act of speaking may typically seem effortless, but it is extraordinarily complex. Speakers must plan the message they wish to convey, choose words and constructions that accurately encode that message, organize those words and constructions into linearly-sequenced utterances, keep track of what has been said, and execute each part of their peaking plants at the correct time. Throughout this process, speakers often face choices in structuring their intended message into an utterance. One central question for a computationally precise theory of language production is thus: When multiple options are available to express more or less the same meaning, what general principles govern a speaker’s choice? To what extent do speakers make choices that potentially facilitate comprehenders, and to what extent do they make choices that are preferable from a production standpoint? Here we approach these questions from the standpoint of contextual predictability, which is known to affect a wide range of speaker choices. Specifically, we investigate the influence of contextual predictability on the encoding of linguistic content manifested by speaker choice in a classifier language, Mandarin Chinese. We report analysis of large naturalistic language datasets and psycholinguistic experimentation.

5PM - 6:15PM New BCS Faculty

Presenter: Prof. Mike Halassa

Talk Title: The frontothalamic system and its roles in cognition

Talk Abstract: Interactions between the cortex and thalamus are known to be relevant to sensation, action and cognition. However, most of our computational and mechanistic understanding of these interactions is in the context of sensory processing. My lab has focused on cortico-thalamic interactions that are directly relevant to cognitive function; cognitive control and flexibility. In my talk, I will show how the thalamus plays previously unrecognized roles in implementing attentional control and working memory, which involve flexible reconfiguration of task-relevant prefrontal representations. Our work leverages circuit-specific manipulations in multiple species, including mice, tree shrews and marmosets.

Presenter: Prof. Ila Fiete

Talk Title: On inferring strongly recurrent circuits from neural activity

Talk Abstract: Understanding neural computation mechanism requires knowledge of the circuitry. Due to the difficulty of mapping complete circuits, there has long been interest in estimating it from simultaneous neural recordings. In this talk, I'll show that connectivity estimates are biased toward inferring connections between unconnected neurons with high correlations, as can happen in strongly recurrent circuits, representing a failure to fully "explain away" even by sophisticated inference algorithms applied to large volumes of data from every node in the circuit. This effect occurs when there is a mismatch between the true network dynamics and the generative model assumed for inference, an inevitable situation when we model the real world. The problem is greatly exacerbated in the typical case of the recorded neurons being a subset of the network, suggesting that connectivity estimates in strongly connected networks should be treated cautiously.

Presenter: Prof. Sasha Rakhlin

Talk Title: Why do Artificial Neural Networks Generalize?

Talk Abstract: Artificial Neural Networks are typically overparametrized and fit the training data exactly (i.e. memorize it). Such a learning mechanism goes against the classical teachings of statistics and machine learning. In this talk, we survey some recent progress on understanding success of memorization (interpolation) methods from the point of view of generalization. In particular, we isolate a phenomenon of implicit regularization, due to high dimensionality and favorable geometric properties of the data, that is partly responsible for success of interpolation as a learning mechanism.

Presenter: Prof. Evelina Fedorenko

Talk Title: The relationship between language and thought

Talk Abstract: Human language surpasses all other animal communication systems in its complexity and generative power. In my group, we use a combination of behavioral, brain imaging, and computational approaches to illuminate the functional architecture of language, with the ultimate goal of deciphering the representations and computations that enable us to understand and produce language. In this talk, I will ask whether language processing shares mechanisms with other aspects of high-level cognition, such as arithmetic, music perception, executive functions, and Theory of Mind. I will present evidence from neuroimaging studies showing that healthy adults strongly engage the brain’s language areas when they understand a sentence, but not when they add and subtract, listen to music, hold information in working memory, or think about another person’s thoughts. I will further talk about individuals with severe aphasia, who, despite their near-total loss of language, are nonetheless able to perform a wide range of complex non-linguistic cognitive tasks. Together, these two complementary lines of evidence demonstrate that many aspects of thought engage distinct brain regions from, and do not depend on, language.