Please join us for a short evening with Mayfield Fund and a few of our portfolio companies, where we’ll spend the first half providing insights from a venture capital firm and what it’s like to work at a startup, then break into a couple of short small-group sessions, providing the opportunity to meet and emerging startups who are hiring Cal students for full-time and internships! Mayfield is...

I will talk about our recent works on computer vision: 1. Smart architecture search methods to discover a highly efficient model called EfficientNet, and EfficientNetV2 which are much more efficient than previous image classification models. 2. Using unlabeled data to improve accuracy and robustness of vision models. The method, named "Noisy Student", requires generating pseudo labels on...

In this talk, we discuss challenges in robot learning in real world including (1) data efficiency, (2) generalization, and (3) long-horizon tasks. Then, we address them by proposing novel frameworks for learning models for planning and acting. We validate the efficacy of our methods on simulated locomotion benchmark, navigation tasks, and real robot manipulation.

The zoom link for this event is https://berkeley.zoom.us/j/95609551998?pwd=bEFlU0ZyUFd6c3VtL2UzUUVqamhMZz09 The cell is not just a small bag with a concentrated solution of proteins and nucleic acids undergoing second order reactions. Cells have polarity, many cellular processes need to be directional, and these cannot occur by simple diffusion between reactants. Instead, the cell uses tiny...

The cloud has become the most common way to deliver commercial software, but it requires building products in a very different way from traditional software, which has not been heavily studied in research. I will explain some of these challenges based on my experience at Databricks, a startup that provides a data analytics platform as a service on AWS and Azure. Databricks manages...

The Satisfiability problem is perhaps the most famous problem in theoretical computer science, and significant effort has been devoted to understanding randomly generated SAT instances. The random k-SAT model (where a random k-CNF formula is chosen by uniformly selecting m clauses from the set of all possible clauses on k distinct variables) is important for several reasons.

In this talk, I will present our efforts towards creating new content in structural image domains from hand-designed fonts to natural complex scenes.

History provides context for the present. Applied to data analysis, past user interactions provide context for present explorations. This thesis investigates different ways to reify user interaction history to help address emerging challenges in the design and programming of data analysis tools. We present three research artifacts that leverage interaction history in different but connected ways.

This talk will cover a three-pronged we employed to co-optimize the algorithms, hardware, and scheduling for deep learning applications.

This spring, Berkeley's Institute of Transportation Studies is hosting a Zoom seminar series on "Applications of Data Science and AI to Equity, Race, and Inclusion in Mobility and Transportation" with the College of Engineering and The Center for Information Technology Research in the Interest of Society and the Banatao Institute. This topic brings a unique and innovative perspective to existing...