The current standard for training brain-computer interface (BCI) machine learning models is user-specific. There is a high interest in developing generic models that are trained on data from other users to minimize BCI calibration time; however, this is limited by noisy, non-stationary brain signals and high inter-user variability. We investigate the trade-off between training data quality and quantity on P300 BCI performance in individuals with amyotrophic lateral sclerosis (ALS) with representative traditional machine learning (stepwise linear discriminant analysis, SWLDA) and deep learning (EEGNet) models. Results show that data quality and domain alignment are more critical than dataset size: user-specific models trained on significantly less data outperformed generic models; generic models trained on ALS data outperformed models trained on non-ALS data; block-averaging of features was mostly detrimental to EEGNet but beneficial to SWLDA; and accounting for inter-stimulus interval differences between ALS and non-ALS data had minimal effect. Our findings highlight the importance of individualized model tuning for reliable P300 BCIs.

We revisit the problem of designing scalable protocols for private statistics and private federated learning when each device holds its private data. Locally differentially private algorithms require little trust but are (provably) limited in their utility. Centrally differentially private algorithms can allow significantly better utility but require a trusted curator. This gap has led to significant interest in the design and implementation of simple cryptographic primitives, that can allow central-like utility guarantees without having to trust a central server.

Flagellated bacteria have been well understood in regards to its adhesion to surfaces and in swimming propulsion. However, its ability to be used as a source of propulsion for artificial microsystems is of great interest to the micro-robotics community; its high efficiency in converting chemical energy to motion is highly attractive for microsystems that demand a low payload and high rate of actuation. In this paper, we describe the behavior of Serratia marcescens bacteria-propelled polystyrene beads in the presence of a chemoattractant, L-threonine. We compare the results of this chemotactic behavior to that bacteria-propelled bead without a chemoattractant. The results from this analysis indicate a clear sign of directionality, as well as an improved bead velocity, for the bacteria-attached microbeads in the presence of a chemoattractant.

For several years, the lectures in our introductory Electrical Engineering and Computer Science (EECS) courses have been videotaped and webcast, mainly as an aid to students with time conflicts that prevent them from attending class. We present the Weiner Lecture Archives - a project to identify, archive, filter, and make available the best of these lectures with their notes on the web. We provide a hierarchical, ontology-driven interface to entire courses, which allows users to choose any topic and/or subtopic to view, from a small snippet of one lecture to one that spans many lectures. Once the topic is chosen, our system launches RealPlayer to play the lecture video in one window while showing synchronized lecture notes or slides in another window. By the spring of 2007, we had finished encoding our department's entire four-course introductory sequence into this system. Student and worldwide use was greater than 100,000 views last year alone, direct feedback has been encouraging, and we hope to expand to other EECS courses in the future.