Cog Lunch: Yudi Xie "Sensory representation mismatch explains working memory capacity limitation" & Thomas Clark "Computationally Characterizing Communication under Constraint in Aphasia"
Description
Speaker: Yudi Xie
Title: Sensory representation mismatch explains working memory capacity limitation
Abstract: The limited capacity of the brain to retain information in working memory has been well-known and studied for decades, yet the root of this limitation remains unclear. Here we built sensory-cognitive neural network models of working memory that perform tasks using raw visual stimuli. Contrary to intuitions that working memory capacity limitation stems from memory or cognitive constraints, we found that pre-training the sensory region of our models with natural images imposes sufficient constraints on models to exhibit a wide range of human-like behaviors in visual working memory tasks designed to probe capacity. Examining the neural mechanisms in our model reveals that capacity limitation mainly arises in a bottom-up manner. Our models offer a principled and functionally grounded explanation for the working memory capacity limitation in terms of sensory representation mismatch. This work highlights the importance of developing models with realistic sensory processing even when investigating memory and other high-level cognitive phenomena.
Speaker: Thomas Clark
Title: Computationally Characterizing Communication under Constraint in Aphasia
Abstract: How do people with impaired language abilities communicate an intended message despite significant constraints? Previous approaches to modeling language production in aphasia have mainly focused on implementational and algorithmic accounts of linguistic deficits, while rational models of language have mainly focused on healthy, neurotypical speakers. Here I discuss work in progress towards a computational-level model of language use in aphasia that seeks to explain observed utterances given a communicative task and a space of possible linguistic constraints. I will propose a conceptual framework for approaching this question, and will discuss possible choices for each component of the model, evaluation metrics, and potential obstacles. Discussion and feedback on this ongoing work are highly welcome.