BYR Achieve · 镜像论坛

Specific suggestions for machine learning people

2008/2/4镜像同步5 回复

Author: Kevin P Murphy @ UBC Big Bull, major works in probabilistic network and related topic Author of BNT(one perfect platform of Bayesian Network) His site has lots of readable codes, kalman filter, conditional random field,... Specific suggestions for machine learning people 1)Learn and use matlab. 2)Join the following emails lists: connectionists and uai. 3)For UBC people: Join the following email lists: lci-lab (ask Valerie to subscribe you), robuddies (Little/Lowe), lm-rg (learning machines: de Freitas), fopi-rg (first-order probabilistic inference: Poole), ea-rg (experimental algorithms: Leyton-Brown/ Hoos). You can add yourself to a mailing list by sending an email to majordomo@cs.ubc.ca with the line "subscribe ListName" in the message body. Also, To be included on our seminars mailing list and notified of all seminars and talks related to Computer Science, please send e-mail to seminars-request@cs.ubc.ca and in the body of your message type: subscribe 4)Recommended reading for machine learning people "All of Statistics", Larry Wasserman, Springer 2004. "Artificial Intelligence: A Modern Approach", Stuart Russell and Peter Norvig, 2nd ed, Prentice Hall 2003 General suggestions 1)Learn and use Latex. (See my Latex tips page.) If you use windows, I recommend the free WinEdt package, which makes LaTeX compilation very easy. 2)Learn and use BibTex, and start building a .bib file for all your references. 3)Learn and use powerpoint. It is the currency of the realm, whether you like it or not. For equations, use the excellent TexPoint plugin. 4)Use CVS for version control. Tortoise CVS is a nice windows interface to this, which is much easier to use than the command line. 5)SVN tutorial for murphyk group. 6)Remember the 80/20 rule (80% of the time is spent in 20% of the code). Hence 80% of your code should be in some high level language, like matlab or Ocaml or R or python. Your time is more valuable than the computer's time. 7)Strive for the goal of reproducible research. This means you store all the code/ parameters you used to generate all the figures in your papers, and you make them publically available. This is not only scientifically honest, but it will help you re-generate old results, eg when it comes time to write a journal version of a conference paper. See Making scientific computations reproducible, by Matthias Schwab, Martin Karrenbach, and Jon Claerbout, Computing in Science & Engineering, Vol. 2, Issue6, Nov.-Dec. 2000, p.61-67. See also Wavelab and reproducible research, Jonathan Buckheit and David Donoho, 1995. 7)For UBC people: learn and use the Wiki

订阅后，新回复会通过你的通知中心匿名送达。