UC Berkeley offers one of the largest programs in EECS in the United States with nearly 1200 undergraduate and 500 graduate students, taught by nearly 90 full time faculties, divided roughly equally between EE and CS. This quantity has been accompanied by quality - the graduate programs in both EE and CS has, in recent surveys, been ranked first in a three way tie with Stanford and MIT. Over the last twenty years, 11 ACM distinguished dissertation Award winners obtained their PhDs at UC Berkeley, more than at any other university.
The faculty-student ratio at Berkeley, a public institution, is not as favorable as at some of the leading private universities, so other factors must play a significant role. Highly selective input is obviously one part of the story. I think another major factor is the emphasis on serious project work in both undergraduate and graduate courses. At the undergraduate level, projects train students to work in teams, a desirable ability for both industry and academia. They also enable students to tackle more ambitious projects, instead of toy problems. At the graduate level, course projects frequently lead to publishable research. Interdisciplinary research is supported by the fact that during course projects, students develop friendships across different subfields of computer science and acquire enough of a working knowledge of other subfields so as to conduct meaningful collaborations.
I will also highlight some other elements of our educational philosophy.

We see a continued convergence of mobile, computer, and consumer electronics industry with rapid advances in smart devices, communications and networking, and new applications and services. New intelligent devices are emerging with powerful 32-bit embedded processors and multi-tasking operating systems. The continued evolution from 2G/2.5G to 3G and advances in PAN/LAN/WAN lead to all-IP infrastructure with high-speed access, multi-radio technology, always-on capability, and seamless connectivity. While voice continues to be a critical driving force for synchronous communications, new data-centric applications, such as messaging, media, push-to-talk, emails, web browsing, location-based service, and corporate data access, create most exciting opportunities for operators, OEM/ODM, developers, consumers, and business.
This talk presents Microsoft’s vision on seamless mobile computing that enables (a) deep connectivity of mobile devices with desktop PCs, backend servers, web, and other devices; (b) automatic detection, seamless roaming and soft handover in a multi-radio environment with the “best” QoS and consistent user experiences; (c) natural user interface with voice dialing, voice command, TTS, ink, and vision; and (d) a powerful platform and ecosystem with compelling applications and services developed by ISVs, OEMs, and operators. I’ll talk about the new advances in the embedded and mobile space, and in particular highlight a plethora of new devices built on Windows CE, PocketPC, and Smartphone platforms. I’ll also touch on a few examples of our active research work on mobility and networking across MSR labs, including seamless roaming, mobile media, navigation, and mesh networks.

This talk will provide an overview of our projects in computer gaming and digital entertainment. For games we will focus on 3D modeling, real-time rendering, and facial animation. Our recent work in these areas include generalized displacement mapping, shell texture, Poisson mesh, iso-charts (for generating texture atlas). Today抯 game platform is going beyond game-playing and becoming a more general platform for digital entertainment. Our research also reflects this emerging trend. For digital entertainment, I will present cool technologies for interacting with digital images and videos, including lazy snapping (an image cut-out tool), Poisson matting, and video toning (for stylizing your home videos). Our technologies are used by game developers all over the world.

The proliferation of networked embedded devices such as wireless sensors ushers in an entirely new class of computing platforms. We need new ways to organize and program them. Unlike existing platforms, systems such as sensor networks are decentralized, embedded in physical world, and interact with people. In addition to computing, energy and bandwidth resources are constrained and must be negotiated. Uncertainty, both in systems and about the environment, is a given. Many tasks require collaboration among devices, and the entire network may have to be regarded as a processor.
The networked embedded computing group at Microsoft Research is developing new architectures, models, and tools for organizing and programming these systems, and innovative applications in areas such as security, transportation, and healthcare. Our goal is to build systems that are easy to use, manage, and program, robust to failures, and secure. We argue that the traditional node-centric programming of embedded devices is inadequate and unable to scale up. We need new service architectures, inter-operation protocols, programming models that are resource-aware and resource-efficient across heterogeneous devices that can range from extremely limited sensor motes to more powerful servers. In this talk, I will describe our on-going work in developing a light-weight, multi-tier architecture that seamlessly mediates information flows between wireless sensors, micro-servers, and PCs, and discuss challenges in resource discovery, sensor tasking, load balancing, and uncertainty management.