TJEN-SIEN LIM AND WEI-YIN LOH & YU-SHAN SHIH. "A Comparison of Prediction
Accuracy, Complexity, and Training Time of Thirty-three Old and New
Regional GIS Specialist
Parks Canada Agency - Ontario Service Centre
lehmann(a)fh-nuertingen.de on 01/23/2002 01:28:17 AM
cc: (bcc: Justin Quirouette/Est-East/PCH/CA)
Subject: [Wekalist] (no subject)
I'm writing my PhD and need some advice....
Has someone a link for a comperison between neural nets and other data-mining
I would like to give the pros and cons for neural nets in data-mining.
Wekalist mailing list
There are new releases of
* the book version (weka-3-0-6)
* the GUI version (weka-3-2-2)
* and the development version (weka-3-3-1)
The Weka Team
I want to ask if somebody on the list have some experiences with compiling
Weka using native code Java compuiler. We have tried to do so with two tools
on Windoews platform: Jove and Excelsior.
The executable code we've obtained was (according to our observations) a
little bit faster (up to 30 % in some I/O operations, less in case of other
funcions). It was also, unfortunately, much less stable than original Java
version. Several times we've got unrecoverable errors or even hangups. That
mostly happened when we tried to process some quite big datasets. In case of
smaller sets the operation of natively compiled WEKA was quite OK.
If anybody has some experience in that, not only in Windoze, please share.
We are really curious if something valuable for ML tools can come from
Marcin S. Szczuka Ph.D.
Institute of Mathematics, Warsaw University
Banacha 2, 02-097 Warsaw, Poland
In a snapshot:
* 5+ years of software development experience.
* At least 2 years of traceable experience/research with one or
more of the following specialized subjects: Information Retrieval/Search
Engine technology, Machine Learning, Data Mining, Collaborative
Filtering, Multi-Agent Systems, Data warehousing and/or OLAP.
* Object/agent-oriented programming is a must.
* Good oral and writing skills.
* Located in the Tri-State area or willing to relocate to New
York at own expense (negotiable)
* Java, and C/C++ proficiency required; XML, Prolog/Lisp and .NET
knowledge is a plus
The ideal candidate must have a minimum of 3 years of Java programming
experience as well as 5-10 years of professional experience as a
programmer. Must have extensive experience developing 3-tier J2EE web
based applications using JSP, servlets, EJB,JMS, XML. The candidate must
have 5-6 years of systems design and implementation as well as
understanding of Object Oriented practices and methodologies. Knowledge
of AI and IR toolkits such as WEKA, SMART, Jakarta-Lucene or MG is a big
plus. UML, WebLogic, Vignette, Interwoven and Lotus Notes are also
desired. Working knowledge of relational databases from Oracle and
Microsoft and experience with software development in financial services
is also strongly desired.
Responsible for the design and implementation of complex 3 tiered
J2EE-based web and desktop applications, writing technical design
specifications, writing programming code, working with QA to test and
debug applications and interfacing with analysts to develop prototypes.
$70k or more (according to experience)
New York, NY - New York
Health Insurance, Life Insurance, Dental Insurance, Disability
Insurance, Paid Vacation, Paid Sick Leave, 401(k), Stock Options
Please contact me directly.
Joaquin Delgado, PhD.
Chief Technology Officer
45 West 25th Street, 9th Floor
New York, NY 10010
This is perhaps a somewhat specialized use of Weka, but I wanted to
mention it all the same. We have found Weka to be a very faithful
companion over the last year or so, and wanted to show how we have been
integerating it into a word sense disambiguation system.
We are happy to announce the availability of the complete source code
distribution for the Duluth systems that participated in the Senseval-2
comparative exercise among word sense disambiguation systems. This is free
software, distributed under the GNU CopyLeft.
This includes a number of components:
SenseTools (v0.1), a suite of Perl programs that convert sense-tagged
text into a feature vector representation suitable for use with the Weka
machine learning system. Users may specify features to be identified in
the text using regular expressions, or features may be automatically
identified using the Bigram Statistics Package (v0.4 or better), which
is also available.
Duluth-Shell, a set of C-shell scripts that tie together the Bigram
Statistics Package, SenseTools, and Weka and should allow a user to easily
replicate the Duluth systems from Senseval-2, and provide a convenient
starting point for further experimentation with corpus-based, machine
learning oriented methods.
You can find SenseTools, Duluth-Shell, the Bigram Statistics Package, and
a pointer to Weka (which was developed at the University of Waikato) at
Please let us know if you have any questions.
# Ted Pedersen http://www.d.umn.edu/~tpederse #
# Department of Computer Science tpederse(a)d.umn.edu #
# University of Minnesota, Duluth #
# Duluth, MN 55812 (218) 726-8770 #