Re: importance of statistics in computer science

From: Ross Clement (clemenr_at_wmin.ac.uk)
Date: 09/22/04 

Date: 22 Sep 2004 02:15:14 -0700

My take on it would be that statistics is very important for Computer
Science, and has two main roles (there are probably others).

Like any science, Computer Science frequently involves the analysis of
data, either from experiments, or from observing computer systems.
Like any science, Statistics is vital for interpreting and analysing
experimental or survey data.

Paul Cohen's book _Empirical Methods for Artificial Intelligence_ is a
good book, and despite the title, I think it would be a useful read
across wide ranges of computer science. I haven't seen the book that
Rich suggests, and hence can't comment on it.

Secondly, statistical methods are frequently used in the design of
computer systems. Choosing an example which is obviously statistics
rather than being based on probability theory, Anna Hart designed a
system for building decision trees (a form of knowledge-based
programming) which made extensive use of chi-square sigificance tests.
Personally I think that statistical, or probability theory, methods
should be used in some cases where fuzzy logic is used at present, but
have no proof of that :-) It's also not uncommon to see "seat of the
pants" algorithms that could be improved if the author was more
familiar with more formal mathematical approaches. An example of this
(though for information theory rather than stats) is that Decision
Trees (at least as they were developed in AI, there was a parallel
development in Stats) originally used 'seat of the pants' heuristics,
but after a suggestion (I can't remember by who) was made to Ross
Quinlin, heuristics based on information theory were adopted.

Cheers,

Ross-c

Relevant Pages

  • Solutions manuals available
    ... An Introduction for the Engineering, ... Understandable Statistics 7th Ed by Charles Henry Brase, ... Automotive Mathematics ... Science of Electronics, The: Digital (Thomas L. Floyd & David ...
    (sci.physics.accelerators)
  • Re: Howard theory - "cannot be computed"
    ... fairytale stories - not science. ... gap distance is the minimum possible gap distance. ... agreed that your statistics would be valid *if* evolution involved ...
    (talk.origins)
  • Re: Kanye West
    ... science for political reasons) that attitude, ... the mind. ... Scientists are human, with all the ... One can tweak statistics to mask a negative result as maybe positive. ...
    (alt.guitar.bass)
  • Re: Democrats try to hide anti-global warming report
    ... also done graduate work in statistics and taught statistics. ... science about something like that in finance, ... post here and his recent E&E paper on Rahmstorf et al ... making adjustments to compensate for that bias. ...
    (rec.gambling.poker)
  • Re: Short-lived professions?
    ... statistics predict the occurrence of improbable events, ... Events remain undetermined and the laws of statistics require ... clear", Josh could know... ... Josh's science has sloughed off the concept that some fences may be ...
    (misc.writing)