Statistical learning algorithms are being used in a wide variety of applications in AI and beyond. Many recent developments fall into one of two broad classes: "probabilistic graphical models," a formalism that exploits the conjoined talents of graph theory and probability theory to build complex models out of simpler pieces, and "kernel methods," a formalism which combines convex optimization with generalized notions of similarity. I provide an introduction to these ideas and I discuss a number of applications: (1) finding bugs in computer programs, (2) gene-finding with DNA from multiple species, (3) automatic image annotation, (4) functional annotation of proteins, (5) learning hierarchical topic models of document collections.(With Alex Aiken, David Blei, Nello Cristianini, Gert Lanckriet, Ben Liblit, Jon McAuliffe, William Noble, Lior Pachter, Martin Wainwright, and Alice Zheng).Beyond the Genome: Current Trends in Computational Molecular Biology--Richard KarpWith the advent of whole-genome shotgun sequencing and the availability of high-throughput technology for measuring macromolecules within cells, the data available to molecular biologists has multiplied and the level of aspiration of the field has been raised. Knowledge about the regulation and expression of genes, the structures and functions of proteins, and the classification of proteins into families is codified in large and growing databases. The central questions now concern systems in which many genes, proteins, protein complexes and small chemical molecules interact to perform cellular functions. These systems are increasingly studied through the lenses of evolution and development: an organism is not considered as a static and isolated entity, but rather through comparison with its relatives in the Tree of Life, and in terms of how its cells have differentiated during embryonic development. The analysis of these systems requires new tools from machine learning, statistics, mathematics and combinatorial algorithms. The speaker will discuss ongoing projects within EECS concerned with the regulation of gene transcription, the identification of protein complexes and regulatory cascades using interspecies comparisons, and the study of embryonic development.Silicon at the End of the Roadmap ... and Beyond--Tsu Jae KingThe integrated-circuit (IC) industry has relied on shrinking transistor geometries for improvements in circuit performance and cost per function for more than three decades. Continued transistor scaling will not be as straightforward in the future as it has been in the past, however, because new materials and non-classical transistor structures will be needed in order to meet the performance specifications of the International Technology Roadmap for Semiconductors. This seminar will begin by describing how new materials and transistor structures may provide a technological pathway for scaling CMOS devices to atomic dimensions. Then, it will give examples of alternative approaches to reducing cost per function, beyond transistor scaling.Intra-Planetary Storage Networks: In search of Ubiquitous, Dependable, and Indestructible Storage--John KubiatowiczIn the past decade we have seen astounding growth in the performance of computing devices. Even more significant has been the rapid pace of miniaturization and related reduction in power consumption of these devices. Based on these trends, many envision a world of ubiquitous computing devices that add intelligence and adaptability to ordinary objects such as cars, clothing, books, and houses. Given this vision, however, one question immediately comes to mind: where does persistent information reside?Berkeley has several related efforts directed at answering this question. One possible answer to this question is OceanStore, a utility infrastructure designed to span the globe and provide continuous access to persistent information. OceanStore builds upon peer-to-peer technologies to provide stable behavior in the face of unstable components. OceanStore data is protected through redundancy and cryptographic techniques. To improve performance, data is allowed to be cached anywhere, anytime. Additionally, monitoring of usage patterns allows adaptation to regional outages and denial of service attacks; monitoring also enhances performance through pro-active movement of data.In this talk, OceanStore will serve as a motivating framework to describe several interesting research efforts here at Berkeley. Naturally, we will discuss the mechanisms of OceanStore and discuss the status of its implementation. We will also discuss the new PetaByte store project, seeking to provide tapeless backup services to EECS as well as other UC campuses through global redundancy. In the course of this talk, we will discuss how self-organizing peer-to-peer technologies are providing vital services to projects such as OceanStore, the PeteByte Store, and database query engines.Universal Access: Computing for Anyone, Anywhere (but not Everyone, Everywhere)--Jennifer MankoffComputers are becoming ubiquitous. Cellphones, PDAs, cruise control, voting machines, the web and the internet represent a cultural and cross-cultural phenomenon that is changing the way we communicate, interact, and work. As researchers, we are tasked with helping to define this new reality. Even more importantly, we must help to insure that it does not create a new class system: haves and have-nots; people who can control and people who are controlled by; people who give up information and entities that gather and use it.At Berkeley, the Group for User Interface Research is exploring these issues. We are developing applications and tools in support of ubiquitous computing. Our work encompasses not only the technology-savvy early adopters so common to the Bay Area, but also people in developing countries, people with disabilities, and end users with minimal computing experience. We are designing and building applications, and investigating overarching issues such as privacy and control.Although we can help to define the direction that universal access will take, the majority of applications will be developed by others. An important challenge, then, is helping others to create truly usable accessible or ubiquitous applications. To this end, we are developing tools and techniques for helping designers to more effectively meet users' needs.
Howe And Sodini Microelectronics
Download Zip: https://ssurll.com/2vIEpw
2ff7e9595c
Comments