1. Clicking ▼&► to (un)fold the tree menu may facilitate locating what you want to find. 2. Videos embedded here do not necessarily represent my viewpoints or preferences. 3. This is just one of my several websites. Please click the category-tags below these two lines to go to each independent website.
source: GoogleTechTalks 53:46 Multi-armed Bandit Problems with Strategic Arms A Google Algorithms Seminar, 4/11/17, presented by Jon Schneider, Princeton University Talks from visiting speakers on Algorithms, Theory, and Optimization 47:50 Computational Aspects of Optimal Information Revelation A Google Algorithms Seminar, 4/14/17, presented by Yu Cheng, USC Talks from visiting speakers on Algorithms, Theory, and Optimization 56:12 Learning from Untrusted Data A Google Algorithms Seminar, 4/414/17, presented by Greg Valiant, Stanford University Talks from visiting speakers on Algorithms, Theory, and Optimization 35:19 Practical Load Balancing with Consistent Hashing A Google Algorithms Seminar, 4/18/17, presented by Andrew Rodland, Vimeo Talks from visiting speakers on Algorithms, Theory, and Optimization 50:21 Hardness of Approximation Between P and NP A Google Algorithms Seminar Talk, 5/18/17, presented by Aviad Rubinstein Google Algorithms Seminar: Talks from visiting speakers on Algorithms, Theory, and Optimization 1:04:59 Assortment Optimization Under a Mixture of Mallows Distribution Over Preferences A Google Algorithms Seminar Talk, 3/28/17, presented by Vineet Goyal "Assortment Optimization Under a Mixture of Mallows Distribution Over Preferences" from the talk series Google Algorithms Semin...
source: GoogleTechTalks 1:00:06 JamBoard: The Future of Collaboration at Google and the Enterprise A Google TechTalk, 4/11/17, presented by Eva Bacon, TJ Varghese, Nick Miceli, Lucy Bourne
ABSTRACT: Googlers spend millions of hours every year in meetings. Google CorpEng seeks to "empower Googler... 54:36 2017 Google Networking Research Summit Keynote Talks A Google TechTalk, 2/7/17, presented by Anees Shaikh, Amin Vahdat, Bikash Koley
ABSTRACT: Opening keynote session for the 2017 Google Networking Research Summit. Talk list:
00:14 Networking Chall... 1:01:34 Large-Scale Distributed Virtual World Systems A Google TechTalk, 4/11/17, presented by Prof. Philip Levis, Stanford University
ABSTRACT: An overview of the dissertations that came out of the Sirikata distributed virtual world project (object l... 51:11 Technology for Today's Conservationist A Google TechTalk, 3/21/17, presented by Steve Schill
ABSTRACT: The ever-increasing pace of technology is providing more tools and information with more detail to more people in more places around ...
source: GoogleTechTalks 2017年3月23日
A Google TechTalk, 3/16/17, presented by Dixon Cleveland
ABSTRACT: Eye-tracking (ET) has become a key means of user interaction and tool to peek into human intention. In this talk, I will cover the theory behind common approaches to ET, the difficulties of effective ET (i.e., why haven't we seen $10 ET everywhere), and the standardization work done by a consortium of industrial and academic partners.
source: GoogleTechTalks 2017年3月3日 A Google TechTalk, 10/28/16, presented by Stacey Turis. ABSTRACT: Stacey Turis is an entrepreneur living with ADHD and giftedness who earned her degree in broadcast journalism from Wichita State University in Kansas. She co-produced and hosted a TV show for a FOX affiliate before pursuing a career in advertising, then graphic design, then market research, then photography, then IT, then acting, then Yoga instruction, then...
In 2012, Stacey self-published the Amazon best-selling memoir, “Here’s to Not Catching Our Hair on Fire”, and also began to manage the now successful Reggae Band, Niu Roots; a job she finally loved enough to stick around for, and still does passionately to this day. She has, through the years, unsuccessfully started twenty-seven businesses and can't remember most of them; though she does remember every lesson those un-successes taught her. She loves speaking to groups of the same kind of wacky folks, where she's not afraid to stop mid-speech and ask, "What was I just talking about?" Stacey lives in Southern California with her husband, two kids, a dog, three cats, and a guinea pig. Other cool things about Stacey I have the largest ADHD Facebook page on the globe - https://www.facebook.com/ADHDSuperhero Best-Selling book on Amazon - https://amzn.com/0983827508 Named Top 10 ADHD Blogger in 2013 by Dr. Oz's Sharecare - https://about.sharecare.com/press-rel... Owner of the largest on-line ADHD Facebook Group Support Page - 9,000 + members
source: GoogleTechTalks 2017年2月16日
A Google TechTalk, 13 Feb 2017, presented by Alexandre Drouin.
ABSTRACT: Antimicrobial resistance is an important public health concern that has implications in the practice of medicine worldwide. Accurately predicting resistance phenotypes from genome sequences shows great promise in promoting better use of antimicrobial agents. For instance, treatment plans could be tailored for specific individuals, likely resulting in better clinical outcomes for patients with bacterial infections.
Sparse machine learning algorithms are appealing tools in this context, since they make use of a concise set of features, which can be further interpreted by domain experts. However, in extremely high dimensional settings, which are common in genomics, the main challenge remains resistance to overfitting.
In recent work, the Set Covering Machine (SCM) algorithm has been used to obtain concise, expert interpretable, models of antibiotic resistance for 6 pathogenic bacterial species. Known and validated resistance mechanisms were recovered within minutes of computation. An empirical benchmark showed that the SCM compared favorably to more complex learning algorithms (e.g., L1-SVM), both in terms of accuracy and sparsity. Moreover, a theoretical analysis of the method revealed that the SCM has an uncharacteristically strong resistance to overfitting in genomic contexts.
In this talk, I will present the SCM algorithm, along with an efficient implementation for genomic data (https://github.com/aldro61/kover). I will rely on theoretical results and an application to antibiotic resistance to demonstrate that this algorithm is well-suited for predictive modeling in genomics."
ABOUT THE SPEAKER: Alexandre Drouin is a PhD candidate in Machine Learning and Computational Biology at the Université Laval, advised by Prof. François Laviolette.
source: GoogleTechTalks 2017年2月10日
A Google TechTalk, 2/2/17, presented by Sriram Murali
ABSTRACT: Imagine living under a sky full of stars. Imagine kids growing up passionate about Astronomy. But, thanks to light pollution, we've lost our connection with the rest of the Universe. Most people living in cities have never seen or believe a sky full of stars exist. So, I made a short film (vimeo.com/178841667) exploring the different levels of light pollution to show people that these stars really do exist. It was a huge success. It was featured on National Geographic and made the news in over 40 countries.
The night skies remind us of our place in the Universe; it makes us feel a connection and gives us identity. We owe it to our future generations to preserve this beauty.
In this talk, I'm going to talk about light pollution, the importance of night skies to mankind, why and how I made this movie, my passion for Astronomy and how each one of you can spread the word and make a better tomorrow.
About the speaker: Sriram Murali is a photographer passionate about Astronomy and the night skies.
Link to Carl Sagan video: https://www.youtube.com/watch?v=5ZHSK...
source: GoogleTechTalks 2017年2月2日 A Google TechTalk, 1/20/17, presented by Brian J. Cardiff ABSTRACT: Crystal is a new programming language that focuses on developer productivity, type safety and execution performance. It is statically checked and compiles to native (machine) code. It combines a type inference algorithm, compile-time macros, compile-time type introspection, automatic union types and Ruby-like syntax, allowing quick prototyping and generating efficient computer programs. It provides a Garbage Collector, uses LLVM as its backend and doesn’t run on a Virtual Machine. In this talk we will show some examples and patterns that arise from combining all these language features. For more information about Crystal: http://crystal-lang.org About the speaker: Brian J. Cardiff Brian is part of the original three-person core team that started Crystal language. He has over 15 years of experience as a professional software developer, most of them at Manas, the company behind Crystal. He has an extensive knowledge on programming languages, which include being a teacher at University of Buenos Aires on Paradigms of Programming Languages for several years. His contributions to Crystal have helped shape its type system and type inference algorithm, one of the key features of the language. Brian also has a keen interest in visualizations and user interaction, which is reflected in the Crystal playground, the built-in interactive code editor that ships with the compiler.
source: GoogleTechTalks 2017年1月9日
A Google TechTalk, 12/15/16, presented by Don Norman and Mickey McManus ABSTRACT: As the number of increasingly intelligent machines increases and their role shifts from automating labor to playing games and exploring concepts, what role will human designers take? Previously, some capabilities were relegated to humans--empathy, intuition, leaps of imagination, and creativity. But as we build intelligent assistants, automatic translators, and self-directed systems, how will we design them--how will we understand them--and how will we study them? Will humans be competitors, collaborators, midwives, or something never before imagined? How can you design with system components which may have unpredictable behaviors?
Join Don Norman and Mickey McManus as they box a few rounds in a lively debate to explore this important and timely topic. About the speakers
Mickey McManus is a pioneer in the field of collaborative innovation, pervasive computing, human-centered design and education. He is a principal of MAYA Design and the chairman of the board.
For over a decade, Mickey served as MAYA’s president, delivering above industry average profit margins—year over year—while consistently re-investing substantial funds back into MAYA’s R&D efforts. These investments form the core of a pool of intellectual property, trade secrets—and most importantly talent—that drives MAYA’s agility, adaptability, and success.
Mickey co-authored Trillions: Thriving in the Emerging Information Ecology (Wiley 2012). The book is a field guide to the future, where computing will cease to be confined to any particular “box,” but instead be freely accessible in the ambient environment.
Don Norman has been a University Professor, a corporate advisor and board member, well-known author and speaker. Most recently he is the Director of the newly established (2014) Design Lab UCSD. He is also co-founder of the Nielsen Norman group and an honorary Professor at Tongji University (Shanghai) in their College of Design and Innovation. He is an IDEO fellow and a member of the Board of Trustees of IIT's Institute of Design in Chicago. Don's latest books are Living with Complexity and The Design of Everyday Things: Revised and Expanded.
source: GoogleTechTalks 2017年1月18日 A Google TechTalk, 12/6/16, presented by Alexey Nezhdanov ABSTRACT: Esperanto is a constructed language that is 100+ years old. I shared my own experience learning this language. Some comments for the video: 0:17 An agitation that the talk has finally started affected my English which is bad enough even without that :( Sorry. 1:45 Complete miss. I meant the country - and a different one - and I would still have been wrong. Zamenhof was born in Białystok which at the time belonged to the Russian Empire (now Poland). 14:18 "Poste" is a proverb, not a pronoun. 19:45 Zurich German kicks in (I'm in the second week of the language course). The correct possessive pronoun would be "mia", not "mis" 21:49 "They" is just a single person that I asked. A Spanish-speaking person said (today) that it looks like Czech language to him. 26:35 The question being asked is "what actual scientific principles were used in creation of Esperanto?" 37:45 The question being asked is "what is an estimate of a chance of meeting another esperantist?" 38:18 The "99.X%" figure that I use is actually my interpretation of the phrase "almost all" plus some other facts 40:30 I misunderstood the question. The question was "is it mostly no problem to split a stream of sounds into individual words" and my answer was to the question "is it easy enough to segment written words into sub-parts". The answer to original question would be "I would expect so, but I don't know".
source: GoogleTechTalks 2016年12月15日 A Google TechTalk, 12/09/16, presented by Matthew O'Connor ABSTRACT: "Engineering Approaches to Combating the Diseases and Disabilities of Aging: Rejuvenating the Mitochondria." This is a talk for a general audience on the work of the SENS Research Foundation to fight age related diseases with a focus on repairing the damage that accumulates as we age. The SENS Research Foundation recently published a paper on their research into repairing cells that lack two of the thirteen essential mitochondrial proteins. The SRF scientists were able to reengineer the two mitochondrial genes and move them to the nucleus of the cell, restoring the missing proteins. This work is significant for both its impact on treating age related diseases but also on childhood diseases resulting from a lack of certain mitochondrial proteins. About the speaker: Matthew was awarded his Master's degree at Northwestern Medical in 1999 for his work studying behavioral neuroscience in aged rodents. In 2005, at Baylor College of Medicine he received a PhD in Biochemistry for his work on proteins that regulate human telomeres. Postdoctoral research includes work at UC Berkeley on muscle stem cells and aging. Since 2010, Dr. O'Connor has headed up the MitoSENS project at the research center in Mountain View, California. His research is focused on "allotopic expression" of mitochondrial genes where his team is engineering mitochondrial genes to be expressed from the nucleus and targeted to the mitochondia. Since 2012 Dr. O'Connor has had broad oversight over many areas of research at SRF. Matthew O'Connor is passionate about performing basic research to combat the diseases and disabilities of aging.
source: GoogleTechTalks 2016年12月13日 A Google TechTalk, 12/5/16, presented by Ulf Adams, Helen Altsuhler, David Stanke. ABSTRACT: Google has more than 2 billion lines of code distributed over more than 9 million source files. This talk will be a deep technical dive into how Google designed a build system to handle that kind of scale. Bazel is Google's platform independent open source build tool, now publicly available in Beta. Bazel has built-in support for both client and server software, including client applications for both Android and iOS platforms. It also provides an extensible framework that you can use to develop your own build rules.
source: GoogleTechTalks 2016年11月16日 The Google Test Automation Conference (GTAC) is an annual test automation conference hosted by Google. It brings together engineers from industry and academia to discuss advances in test automation and the test engineering computer science field. It is a great opportunity to present, learn, and challenge modern testing technologies and strategies. The first GTAC was held at the Google London office in 2006. GTAC 2016 will be held on Nov 15-16 at Google Sunnyvale.
source: GoogleTechTalks 2016年11月15日 The Google Test Automation Conference (GTAC) is an annual test automation conference hosted by Google. It brings together engineers from industry and academia to discuss advances in test automation and the test engineering computer science field. It is a great opportunity to present, learn, and challenge modern testing technologies and strategies. The first GTAC was held at the Google London office in 2006. GTAC 2016 will be held on Nov 15-16 at Google Sunnyvale.
source: GoogleTechTalks 2016年11月10日 A Google TechTalk, 10/21/16, presented by Boris Altshuler. ABSTRACT: Quantum Computers (QC) consist of a large number of interacting quantum bits. Solutions of computational problems are encoded in bit-strings which result from problem-specific manipulations. In contrast with Classical Computers, the state of a QC is characterized by a quantum superposition of the bit-strings (a wave function) rather than by a particular bit-string representing a computational basis. Instead of usual focus on quantum algorithms, here we will discuss QC using concepts from many-body physics as quantum dynamical systems. Recent progress in understanding the dynamics of quantum systems with large number of degrees of freedom is based on the concept of Many-Body Localization: the eigenstates can be localized in the Hilbert space in a way similar to the conventional real space Anderson Localization of a single quantum particle by a quenched disorder. Depending on the temperature (total energy) or other tunable parameters the system can find itself either in the localized or in the many-body extended phase. In the former case, the system of interacting quantum particles/spins cannot be described in terms of conventional Statistical Mechanics: the notion of the thermal equilibrium loses its meaning. Moreover the violation of the conventional thermodynamics does not disappear with the Anderson transition to an extended state. In a finite range of the tunable parameters we expect the non-ergodic extended phase: the many-body wave-functions being extended are multifractal in the Hilbert space making thermal equilibrium unreachable in any reasonable time scale. It means the system by itself keeps some memory of its original quantum state. This property can be extremely useful for quantum computation, which cannot be implemented without connection between the remote parts of the Hilbert space, i.e. states localized in the computational basis are useless. The ergodic states should also be avoided: in the Hilbert space of high dimension they easily lose the quantum information. We will discuss evidences for the existence of delocalized non-ergodic systems and speculate about their properties by comparing them with non-integrable classical dynamical systems such as Solar Systems.
Speaker Info: Boris Altshuler works in the field of Condensed Matter theory. He made substantial contributions to the understanding of the effects of disorder, quantum interference and interactions between electrons on the properties of bulk, low-dimensional, and mesoscopic conductors. Boris was educated in Russia. He graduated from the Leningrad (now St. Petersburg) State University and joined Leningrad Institute for Nuclear Physics first as a graduate student and later as a member of the research stuff. His PhD thesis advisor was Arkadii Aronov. After moving to USA Boris was on faculty of the Massachusetts Institute of Technology and later of the Princeton University. He was also a Fellow of NEC laboratories America (Princeton, NJ). Now he is a professor of Physics at Columbia University. Boris Altshuler is a recipient of a number of scientific awards - the most significant are 1993 Hewlett-Packard Europhysics Prize (Agilent Prize) and 2003 Oliver Buckley Prize of American Physical Society. He is a member of the National Academy of Sciences and of the American Academy of Arts and Sciences. He is also a foreign member of The Norwegian Academy of Science and Letters and of the Academy of Romanian Scientists.
source: GoogleTechTalks 2016年11月10日 A Google TechTalk, 10/6/16, presented by Garnet Chan ABSTRACT: Quantum mechanics is the fundamental theory underlying all of chemistry, materials science, and the biological world, yet solving the equations appears to be an exponentially hard problem. Is there hope to simulate the quantum world using classical computers? I will discuss why simulating quantum mechanics is not usually as hard as it first appears, and give some examples of how modern day quantum mechanical calculations are changing our understanding of practical chemistry and materials science.
Speaker Bio: Garnet Chan recently joined the Cal Tech faculty as the Bren Professor in Chemistry. Before that he was the A. Barton Hepburn Professor of Chemistry at Princeton University, where he was also a member of the physics faculty. Professor Chan received his PhD from the University of Cambridge in 2000. He was born in London and grew up in Hong Kong. Professor Chan's research lies at the interface of theoretical chemistry, condensed matter physics, and quantum information theory, and is concerned with quantum many-particle phenomena and the numerical methods to simulate them. Over the last decade, his group has contributed to and invented a variety of methods addressing different aspects of quantum simulations, ranging from the challenges of strong electron correlation, to treating many-particle problems in the condensed phase, to dynamical simulations of spectra and coupling between electron and nuclear degrees of freedom. Some of these methods include density matrix renormalization and tensor network algorithms for real materials, canonical transformation-based down-foldings, local quantum chemistry methods, quantum embeddings including dynamical mean-field theory and density matrix embedding theory, and new quantum Monte Carlo algorithms. The primary focus is on methodologies for problems which appear naively exponentially hard, but where an understanding of inherent physics, for example in terms of the entanglement structure, allows for calculations of polynomial cost.
source: GoogleTechTalks 2016年9月15日
A Google TechTalk, 9/7/2016, presented by Professor Aarne Ranta, University of Gothenburg.
Speaker's errata:
4:57: “sixteen forms” should be “twenty-six”
19:32: “more than 2000 members” should be “200” as on the slide
ABSTRACT: GF (Grammatical Framework) is a grammar formalism that was first released at Xerox Research in 1998 and later became an open-source collaborative project. GF is thus at least a decade younger than the major grammar formalisms (LFG, HPSG, TAG, CCG) and has grown up in an era when computational linguistics is dominated by statistical methods rather than grammars. Its background is in fact quite different from the major grammar formalisms, as its roots are in theorem provers and compiler construction rather than theoretical linguistics.
The original mission of GF was to make it easy to implement multilingual controlled language systems, where a semantic interlingua serves as a hub between multiple languages. In such a system, translation works as parsing the source language into an interlingua followed by generation into the target language. Unlike in many other interlingual systems, the interlingua is not fixed but can be easily changed e.g. to adapt to application domains. Thus GF has been used to implement software specification systems, spoken dialogue systems, mathematical teaching tools, tourist phrasebooks, and many other applications, in which up to 30 parallel languages are involved.
In recent years, GF has also scaled up to wide-coverage parsing and translation, resulting for instance in the mobile app GF Offline Translator. While not quite as good in open-domain tasks as state-of-the-art statistical systems, the GF translator has some advantages: compact size (15 languages available offline in 30 megabytes), inspectability (via syntax trees and other grammatical information), and domain-adaptability. The traditional weakness of grammars, their labour intensiveness, is relieved by software techniques that make the development of grammars in GF orders of magnitude faster than with traditional methods.
Another emerging usage of GF is dependency parsing. The booming initiative of Universal Dependencies (UD) has turned out to be very similar to the interlingua used in the wide-coverage GF translator, so that GF trees can be automatically converted to UD trees. Since GF trees support generation in addition to parsing, the mapping makes it possible to bootstrap UD treebanks for new languages. More generally, the use of UD data in combination with GF grammars suggests a way to build hybrid systems that combine data-driven UD parsing with the precise semantic analysis and generation of GF.
ABOUT THE SPEAKER: Aarne Ranta is Professor of Computer Science at the University of Gothenburg. He defended his PhD at the University of Helsinki in 1990. After seven years as Junior Fellow of the Academy of Finland, he worked at Xerox Research Centre Europe in Grenoble in 1997-1999, starting the development of Grammatical Framework (GF), after which he joined the Department of Computing Science of Chalmers University of Technology and University of Gothenburg. Ranta’s research interests have covered type theory, functional programming, compiler construction, and, as his main field, computational linguistics. His has followed the mission to formalize the grammars of the world and make them available for computer applications. In this work, he has been helped by 10 PhD graduates and by a community of over 200 GF contributors. Ranta is currently on a partial leave from the university to work for the start-up company Digital Grammars AB, which develops reliable language technology for producers of information.
# click the up-left corner to select videos from the playlist
source: GoogleTechTalks 2016年10月20日
Sessions of the 2016 Adiabatic Quantum Computing Conference held at Google's Los Angeles office from June 26-29, 2016
source: GoogleTechTalks 2016年10月12日
A Google TechTalk, 10/5/16, Presented by Anirudh Sivaraman
ABSTRACT: The evolution of network routers and switches has been driven primarily by performance. Recently, thanks in part to the emergence of large data centers, the need for better control over network operations, and the desire for new features, programmability of switches has become as important as performance. In response, researchers and practitioners have developed reconfigurable switching chips that are performance-competitive with line-rate fixed-function switching chips. These chips provide some programmability through restricted hardware primitives that can be configured with software directives.
This talk will focus on two abstractions for programming such chips. The first abstraction, packet transactions, lets programmers express packet processing in an imperative language under the illusion that the switch processes exactly one packet at a time. A compiler then translates this sequential programmer view into a pipelined implementation on a switching chip that processes multiple packets concurrently. The second abstraction, a push-in first-out queue, allows programmers to program new scheduling algorithms using a priority queue coupled with a program to compute each packet's priority in the priority queue. For the first time, these two abstractions allow us to program several packet-processing functions at line rate. These packet-processing functions include in-network congestion control, active queue management, data-plane load balancing, network measurement, and packet scheduling.
This talk includes joint work with collaborators at MIT, Barefoot Networks, Cisco Systems, Microsoft Research, Stanford, and University of Washington.
SPEAKER BIO: Anirudh Sivaraman is a PhD student at MIT working in the Computer Science and Artificial Intelligence Laboratory (CSAIL). Before coming to MIT, he was an undergraduate in the Computer Science and Engineering department at IIT Madras and graduated with a BTech in 2010.http://web.mit.edu/anirudh/www/
# click the upper-left icon to select videos from the playlist
source: GoogleTechTalks 2016年9月6日
A Google TechTalk, July 23, 2016, presented by Alexander Thiemann
ABSTRACT: This talk will give an insight on web development using the Haskell web framework «Spock» (https://www.spock.li ). It will give a short overview of the possibilities like type-safe routing, hyperlinks, sessions, database queries, templates and JSON parsing/generation and then we will also look at a technique to build a full stack application with Spock and GHCJS as used in production in TramCloud. https://wiki.haskell.org/ZuriHac2016
53:46 Multi-armed Bandit Problems with Strategic Arms A Google Algorithms Seminar, 4/11/17, presented by Jon Schneider, Princeton University
47:50 Computational Aspects of Optimal Information Revelation A Google Algorithms Seminar, 4/14/17, presented by Yu Cheng, USC
56:12 Learning from Untrusted Data A Google Algorithms Seminar, 4/414/17, presented by Greg Valiant, Stanford University
35:19 Practical Load Balancing with Consistent Hashing A Google Algorithms Seminar, 4/18/17, presented by Andrew Rodland, Vimeo
50:21 Hardness of Approximation Between P and NP A Google Algorithms Seminar Talk, 5/18/17, presented by Aviad Rubinstein
1:04:59 Assortment Optimization Under a Mixture of Mallows Distribution Over Preferences A Google Algorithms Seminar Talk, 3/28/17, presented by Vineet Goyal
