oktober 09, 2007

Lite länkar till videoföreläsningar 20071009

En dump av videoföreläsningar sparade i Bloglines. Allt har inte setts men verkar skoj av en eller annan orsak.


* Decision Science News R video tutorial number 2.


* UCTV Game Theory 2007. Kurs i spelteori. Se även kursens outline.


* En samlig av vetenskapliga experiment av mer vardagligt slag finns på Robert Krampf's Science Videos.


* KDD 2007 The 13th International Conference on Knowledge Discovery and Data Mining (en massa föreläsningar). För den delen finns det flera andra föreläsningar på videolecures.net såsom AAAI-07 AI Video Competition och PASCAL Bootcamp in Machine Learning.


* David Henderson, Charles Hooper Making Great Decisions


The phrase "work smarter, not harder" has been repeatedly ridiculed in Dilbert and ... all » elsewhere, not because it is a poor idea, but because it is thrown like a brick lifesaver to drowning employees. It's like telling someone to be happier, healthier, and richer. What people need is a plan for doing so.

In "Making Great Decisions" the authors show readers how to achieve their objectives. They offer a better way to look at problems so that solutions are easier to find.

Speaker: David R. Henderson, Ph.D. David R. Henderson is an economics professor at the Naval Postgraduate School in Monterey and a research fellow with the Hoover Institution at Stanford. He was a senior economist with President Reagan's Council of Economic Advisers.

Speaker: Charles L. Hooper Charles L. Hooper is President and co-founder of Objective Insights, Inc., a consulting firm dedicated to providing health care companies with marketing and financial analysis to help them make informed decisions about their business opportunities.


* Ross Anderson Searching for Evil

Computer security has recently imported a lot of ideas from economics, psychology and ... all » sociology, leading to fresh insights and new tools. I will describe one thread of research that draws together techniques from fields as diverse as signals intelligence and sociology to search for artificial communities.

Evildoers online divide roughly into two categories - those who don't want their websites to be found, such as phishermen, and those who do. The latter category runs from fake escrow sites through dodgy stores to postmodern Ponzi schemes. A few of them buy ads, but many set up fake communities in the hope of having victims driven to their sites for free. How can these reputation thieves be detected?

Some of our work in security economics and social networking may give an insight into the practical effects of network topology. These tie up in various ways with traffic analysis, long used by the signals intelligence agencies which trawl the airwaves and networks looking for interesting targets. I'll describe a number of dubious business enterprises we've unearthed. Recent advances in algorithms, such as Newman's modularity matrix, have increased the robustness of covert community detection. But much scope remains for wrongdoers to hide themselves better as they become topologically aware; we can expect attack and defence to go through several rounds of coevolution. I'll therefore end up by talking about some strategic issues, such as the extent to which search engines and other service providers could, or should, share information in the interests of wickedness detection.

Speaker: Ross Anderson Ross Anderson is one of the top security researchers in the world.


* Philip Chan Learning Rules for Anomaly Detection

Anomaly detection has the potential to detect novel attacks, however, keeping the false ... all » alarm rate low is a challenging task. We discuss the LERAD algorithm that can learn concise and accurate rules for anomaly detection and demonstrate its effectiveness in network and host datasets. We will also discuss our recent work (KDD 07) on weighting versus pruning during the rule validation.

If there is more time, I can also talk about:

As mobile devices become more pervasive, we study the problem of spatial-temporal anomaly detection for identifying potential abuse. We discuss the STAD algorithm and show its performance on a cell phone dataset.


* Geoffrey West Scaling Laws In Biology And Other Complex Systems


Life is very likely the most complex phenomenon in the Universe manifesting an ... all » extraordinary diversity of form and function over an enormous range. Yet, many of its most fundamental and complex attributes scale with size in a surprisingly simple fashion. For example, metabolic rate (the power required to sustain the system) scales as approximately the 3/4-power of mass over 27 orders of magnitude from molecular levels up to the largest multicellular organisms. Similarly, time-scales, such as lifespans and growth-rates, increase with exponents which are typically simple powers of 1/4. It will be shown how these universal quarter-power scaling laws follow from fundamental generic principles embedded in the dynamics and geometry of underlying networks, leading to a general quantitative theory that captures essential features of many diverse biological systems. Examples will include animal and plant vascular systems, growth, cancer, aging and mortality, sleep, DNA nucleotide substitution rates. These ideas will be extended to discuss social organisations such as cities and firms: to what extent, if at all, can we think of these as very large organisms and therefore as an extension of biology? Analogues to metabolic rate and behavioral times in cities scale counter to their behaviour in biology. Driven by innovation and the creation of wealth this has dramatic implications for their growth, development, sustainability and pace of life which, left unchecked, potentially sow the seeds for their collapse.

Geoffrey West is a theoretical physicist whose primary interests have been in fundamental questions in physics, especially those concerning the elementary particles, their interactions and cosmological implications. Prior to joining the Santa Fe Institute as a Distinguished Professor in 2003, he was the leader, and founder, of the high energy physics group at Los Alamos National Laboratory, where he is one of only approximately ten Senior Fellows. «


* Lynn Robertson Is That My Brother? Perceptual and Neurobiological Factors in Face Blindness


Face blindness (technically known as prosopagnosia) is a condition in which people with otherwise normal vision cannot discriminate one ... all » face from another. They may not be able to pick out their own husband or children in a crowded room or even themselves in a mirror. One woman reported she once had to crinkle her face in a crowded rest room to discriminate herself from others in the mirror. This problem can occur through injury to particular areas within the brain (either through head trauma, stoke or surgery), but it can also occur developmentally. In the latter case, the brain appears completely normal, yet developmental prosopagnosics (DP) have never learned to accurately discriminate faces.

There is a large scientific body of work on face perception published in the psychological, social and neurobiological literature, and I will highlight some of the more important findings. I will then discuss work from my own laboratory on perceptual processing of faces; emphasizing training methods we have developed to help individuals with DP identify faces, sometimes for the first time in their lives. This discussion will be complemented by inclusion of documented neurobiological and cognitive changes that accompany the emergence of face recognition abilities.


* Erin McKean Wordmaking: What it take to succeed in hacking English and invent a new word


Earn the basics of word formation in English, get "raw materials" for new words, and invent ... all » your own word (and have it critiqued) before you let it loose into the English language. The maker of the "best new word" (as voted on by the participants) will win a new dictionary.

Posted by hakank at 06:53 EM Posted to Diverse vetenskap | Video podcasts | Comments (2)

september 26, 2007

Videoföreläsning om R (del 1, en kort introduktion)

Decision Science News (en av favoritbloggarna kring beslutsteori, dataanalys etc) har nu börjat med videoföreläsningar i dataanalysverktyget R. Se vidare R video tutorial number 1. Det är en kort introduktion men kan vara början på något trevligt.

Som presentatören säger så rekommenderas en bok (eller två) i R (eller det mycet snarlika språket S). En förhoppningsvis fullständig lista över relaterare böcker finns här. Några av mina favoriter är:
* William N. Venables, Brian D. Ripley. S Programming (ISBN: 9780387989662). En introduktion för den som vill gå in på djupet i själva R-språket (eller S).
* William N. Venables, Brian D. Ripley: Modern Applied Statistics with S, 2002 (ISBN: 9780387954578). Det är en relativt avancerad bok men innehåller mycket matnyttig information, många tips och funktioner.

Se även Några videoföreläsningar om statistik, sannolikhet och data mining.

Posted by hakank at 10:31 EM Posted to Statistik/data-analys | Video podcasts | Comments (1)

juli 23, 2007

Terence Tao videoföreläsning 'Structure and Randomness in the Prime Numbers' (matematik, talteori)

Terence Tao anses vara en av matematikens klarast lysande stjärnor. Enligt artikeln Terence Tao: The "Mozart of Math":


"Terry is like Mozart -- except without Mozart's personality problems," said John Garnett, professor and former chair of mathematics in the UCLA College. "Mathematics just flows out of him." "Mathematicians with Terry's abilities appear only once in a generation," said Garnett. "He's probably the best mathematician in the world right now. Terry can unravel an enormously complicated mathematical problem and reduce it to something very simple. We're amazingly lucky to have him at UCLA."


Hans videoföreläsning Structure and Randomness in the Prime Numbers (.smil videoformat) är en populärt hållen och mycket trevlig föreläsning om talteori, t.ex. primtalstvillingar och Riemann-hypotesen (som anses vara ett av de knepigaste matematiska problemen). Slides (PDF).

(Not: Jag hade lite problem med att videoströmmen slutade att strömma efter cirka 8 minuter och efter cirka 35 minuter. Det vara bara att starta om på nytt. Ibland funkade det att snabbspola, men inte alltid.)


Se även
Taos blogg What's new (som ersätter den gamla "proto-bloggen" What's new)
Information for media, en samling av artiklar om Tao
En kort TV-biografi (strömmande bilder, alltså)
Några kapitel (.ps-fil) från boken Solving Mathematical Problems: A personal perspective. Och kan beställas t.ex. här (ISBN: 9780199205608).


(Via den relativt nyfunna favoritbloggen God Plays Dice.)

Posted by hakank at 06:18 EM Posted to Matematik | Video podcasts | Comments (1)

juli 03, 2007

Några videoföreläsningar om statistik, sannolikhet och data mining

Semester och det blir naturligtvis lite "hängmatte", t.ex. i form av videoföreläsningar.

Så här är några föreläsningar att förgylla sommaren med:

Davis Mease: Statistical Aspects of Data Mining
David Mease om "Statistical Aspects of Data Mining" med R (www.r-project.org) och Excel. Mease har detta även som en "riktig" (IRL inte bara URL) kurs med kurssajten www.stats202.com. Det är lugnt tempo som går igenom grunderna i både teori och verktyg.

Statistical Aspects of Data Mining (Stats 202) Day 1
Statistical Aspects of Data Mining (Stats 202) Day 2


I skrivande stund har endast publicerats ovanstående två föreläsningar. Efterkommande borde dyka upp via denna sökning på video.google.com .

Kursbok är Tan, Steinbach, Kumar: Introduction To Data Mining (ej läst).


Peter Donnelly: How juries get fooled by statistics
Peter Donnelly How juries get fooled by statistics. Visar flera av våra ointuitioner som finns inom området, ofta underhållare på ett traditionellt brittiskt-Grants sätt. Finns även på TEDTalks här.
Beskrivning:

Oxford mathematician Peter Donnelly explores the common mistakes we make in interpreting statistics, and the devastating impact these errors can have on the outcome of criminal trials. Statistical uncertainty and randomness, he says, confound many of our assumptions about the world. He shares the case of a British woman wrongly convicted of murdering her two infants -- a verdict reached, in part, by the misuse of statistics.

[För den som vill läsa mer t.ex. om det där mynt-experimentet finns en relativt djupgående redogörelse i Anirban DasGupta: Sequences, Patterns and Coincidences (PDF).]


Brian Brushwood: Scams, Sasquatch, and the Supernatural
Brian Brushwood Scams, Sasquatch, and the Supernatural. Är mest om debunkning av olika typer av pseudo-vetenskap, men innehåller en akt om sannolikheter (sammanträffanden). Underhållande och med högt tempo.


Ever wonder how those guys on TV seem to talk to the dead? What about ESP and psychic surgery? How do street scams and cons work? Want to ... all » know how YOU can trick your friends into believing you have psychic powers?

As a magician, Brian’s wise to all the tricks used by frauds, tricksters, and con artists …and now he’s ready to take YOU to scam school. This is no ordinary lecture: we’re talking hands-on experiments, a live performance of psychic surgery, free giveaways of cash and prizes, and all the secrets TV psychics DON’T want you to know.

Topics covered include: scams, cons, ESP, UFOs, skeptic, skepticism, dowsing, astrology, memory, alternative medicine, psychic surgery, pseudoscience, coincidence, and crop circles.

Relaterat

Persi Diaconis On Coincidence (som jag iofs skrev om för några år sedan men rekommenderar gärna igen).

Talks Hans Rosling: New insights on poverty and life around the world
Talks Hans Rosling: Debunking third-world myths with the best stats you've ever seen


En kul semesterövning är att analysera lottodata från det sydafrikanska lotteriet. Se vidare bloggen Freakonomics: Is the South African Lottery Rigged? A Hands-On Exercise for Bored Blog Readers.

Uppdatering
Äh, jag glömde ju Scott McClouds grafiska novella The Right Number:


The Right Number is a projected three part online graphic novella about math, sex, obsession and phone numbers presented in an unusual zooming format. Click above to read Parts One and Two. (Part Three will hopefully be completed and available before too long.)

(Not: Serien är skriven 2003 så frågan är om det kommer en tredje del. De två första delarna är dock ganska självständiga. Via EconLog.)




(Avslutningsvis - men egentligen helt orelaterat till ovanstående - kan noteras att det för några veckor sedan var 4-årsdagen av denna bloggs födelse, som inträffade helt utan några virtuella tårterier. Referens till tidigare bemärkelseskriverier finns i förra årets 3-årsdag samt en drapa om varför bloggar bör ses som mer än en dagbok.)

Posted by hakank at 10:56 FM Posted to Sammanträffanden | Statistik/data-analys | Video podcasts

juni 21, 2006

Martin Nowak: Why we cooperate – videoföreläsning

Videoföreläsningen Why we cooperate är en populärt hållen översikt kring hur samarbete inom biologi och samhälle kan uppstå och frodas, kopplat till diskussion om evolutionär spelteori. T.ex. beskrivs det upprepade fångarnas dilemma och dess evolutionärt bästa strategi. Nyckelorden för (mänskligt) samarbete är reciprocitet, rykte (reputation) och rättvisa (fairness).

Föreläsaren är Martin Nowak, Harvard University som har skrivit en hel del inom ämnet.

Not: Nowak pratar rätt fort. Det ord som framförallt i början låter som "mismetical" är naturligtvis "mathematical". Kan kanske vara bra att veta.


Se även
Evolutionary Dynamics är en PDF-presentation som till stora delar är samma som för föreläsningen ovan.

en.wikipedia: Prisoner's dilemma
sv.wikipedia: Fångarnas dilemma.
Stanford Encyclopedia of Philosophy: Evolutionary Game Theory

Nya rön i fångarnas dilemma
Nya rön i fångarnas dilemma - lite länkar
Lite mer om trust/reputation samt peer-to-peer

VirtualLabs in evolutionary game theory, simuleringar kring evolutionär spelteori. I finns även några relevanta simuleringar/experiment.

Bok: Game Theory Evolving - A Problem-centered Introduction to Modeling Strategic Interaction av Herbert Gintis.


(Länken till föreläsningen via The Bumble Bee Ken Thompson's shared know-how on team dynamics, virtual collaboration, emerging technology and bioteaming.)


Andra bloggar om: , , , .

Posted by hakank at 08:19 EM Posted to Agentbaserad modellering | Spelteori och ekonomi | Video podcasts

Pascal Virtual Playground: machine learning videoföreläsningar

(Mer videos fast denna gång något mer akademiskt än Lite nördmusikmotvideos.)

PASCAL Virtual Playground innehåller ett stort antal videoföreläsningar kring machine learning statistisk modellering och liknande ämnen. För att se föreläsningarna synkroniserade med presentationerna krävs Internet Explorer, men det finns även möjligheter att se de allra flesta videoföreläsningarna i andra webläsare via "Preview video" (då utan synkroniserade presentationer) .

T.ex. Machine Learning Summer School 2005 - Canberra.

PASCAL står för Pattern Analysis, Statistical Modelling and Computational Learning och är ett nätverk för att samla forskare inom forskningsområdet. Från dess hemsida


The objective is to build a Europe-wide Distributed Institute which will pioneer principled methods of pattern analysis, statistical modelling and computational learning as core enabling technologies for multimodal interfaces that are capable of natural and seamless interaction with and among individual human users.

Se även
PASCAL EPrints


(Via Machine Learning Thoughts.)


Andra bloggar om: , , .

Posted by hakank at 07:49 FM Posted to Machine learning/data mining | Video podcasts