Logo Home   Documents and Lists   Create   Site Settings   Help    Up to /
Icon
2005 Machine Learning Class
Home
 
 
 
Documents
Pictures
Icon Class Pictures
Lists
Discussions
Surveys
Icon
New New Discussion
|
Filter Filter
 
Subject
Posted By
Modified
 
upload
 
If you divide your big file into some smaller ones, you can upload them separately.
 
 
Tao HONG
12/13/2005 12:08 AM
 
Tao Upload Problem
 
Tao:
I think the problem may be the size of everything. I got fed up and just uploaded it to my NCSU FTP space and provided a link. You can use the following tool to create a www folder with the right permissions:
http://www.ncsu.edu/it/essentials/web_pages/www4_setup.html
Tejas PATEL
12/13/2005 12:06 AM
 
Tejas Patel's Supporting Files
 
Tejas PATEL
12/12/2005 11:46 PM
 
re: upload the file
 
I've tried three computers and two kinds of explorers, but still can not upload..
Tao HONG
12/12/2005 11:34 PM
 
I haven't yet run into that problem. You might try another browser or computer...??
 (no text)
Mark White
12/12/2005 10:49 PM
 
Uploading Supporting Files
 
Class:
Every time I try to upload my supporting files, the submission times out and I get a web page busy error. This take a couple of minutes. Has anyone else had this problem, and if so, how do I get around it?
Thanks!
Tejas PATEL
12/12/2005 10:33 PM
 
Great Question: Yet another possibility...
 
Also, perhaps in combination with my previous posting, you might want to try using a 50% or 30% size training set (e.g., this is one of the "bag-learner" parameters; also you can use filters to perform such sub-sampling) INSTEAD of the full 100% or very small 10% that I talked-about in class.
Mark White
12/10/2005 8:57 PM
 
Great Question: Here's another idea that may or may not work depending on dataset, etc.
 
Use a very "high-prunning" criteria for J48, e.g. 0.05 or 0.01 or higher. Thus, only the nodes with a large number of examples substantiating there validitity are included in the final tree-model that is output by the J48 learner. In other base-learners, there is often something equivalent, even in non-tree learners. It's equivalent in the sense that only very "certainly valid"...
Mark White
12/10/2005 9:44 PM
 
Yes, it is Monday night. PLEASE, PLEASE view the project schedule and DETAILED DIRECTIONS at the Bottom of the homepage (HERE!).
 (no text)
Mark White
12/10/2005 8:31 PM
 
Non-lopsided experts
 
The filtering that Dr. White explained the last day of class trained the two experts in a kind of lopsided fashion. The first base learner was trained on the correctly classified instances of an identical J48 and the second base learner was trained on the leftovers. Anyone have any interesting ideas on how to more evenly create experts? One idea is to sort of bias each base learner by only giving...
Taylor BOOTH
12/10/2005 3:59 AM
 
Paper Due Date
 
I'm pretty sure it's at 11:55p on the 12th (Monday).  Bummer, eh?  Good luck to all!
Evan MCCAULEY
12/9/2005 8:29 PM
 
Paper Due Date
 
I dug through the notes and the website, but I can't seem to find the due date for the paper.  Does anyone know?
Tim WRIGHT
12/9/2005 8:01 PM
 
To Tejas and Jason
 
Thanks a lot for both of your reviews.
Regarding the question ("relative size of evaluation size") that Tejas raised in my paper, it is from Dr. White's lecture note 21, page 6 and page 7.
It's also the same question I find in your papers. Seems that you were using "66%" for all the experiments. Probably "66%" is the default setting in Weka...anyway, I'm not sure why...
Tao HONG
12/7/2005 11:03 PM
 
Random Forest Doesnt nessary use Un-Weighted Voting!
 
A couple papers I reviewed mentions that the Random Forest uses only unweighting voting. This is not entirly true. In Weka all the trees have equal votes. In the True Brehman algorithm each tree can have different weights or they could have equal weights depending on how you set up the experiment.
 
Fred
Fred LIVINGSTON
12/7/2005 8:07 PM
 
Report Formats
 
In the papers that I have reviewed and some of the others that I have peeked at very few are formatted in the IEEE report format.  Some may find this document very helpful. 
Attachment
Jonathan SIMMONS
12/7/2005 3:08 PM
(Messages 1 to 15) Next Next 
New New Document
|
Upload Upload Document
|
New Folder New Folder
|
Filter Filter
|  
TypeNameProjectDescriptionFilter
Icon
KStar Paper
AggregationK*: An Instance-based Learner Using an Entropic
Distance Measure
John G. Cleary,
Leonard E. Trigg,
Dept. of Computer Science,
University of Waikato,
New Zealand.
e-mail:{jcleary,trigg}@waikato.ac.nz

Information about the K* algorithm
Icon
Results
AggregationExploration of impact of master training set size, training set percentage, and number of random trees.
Icon
Experiments
AggregationThis file includes some modified KOML files, the experiments results(*.csv files, *xls files) and a short summary.

Tao Hong
Icon
Updated_Results
AggregationThe graphs are relalbelled, Thanks to Dr. White. There are also some new results. Generally the excel sheet having the data wouldn't make sense just looking at them. Graphs are better representation any time.
Icon
OutputExcelFilesWithGraphs
AggregationThis file includes the compiled Excel files which also has all the graphs of comparison. Compared Voting with different parameter changes. Compared Meritocracy with different Meta leraners and more parameter changes.
Icon
SomeMoreXMLFiles
AggregationHere are some more XML files for Voting(democracy) as well as Meritocracy experiments. I have done some experiments with Stacking C with linear regression. This also includes those files. I am uploading some results file in the next post.

Rachana
Icon
WekaInJava
AggregationThis will allow you to run Weka and analyze your CSV files in Eclipse.  It also contains all the files that I used to run my experiments (Conference Paper.doc).
Icon
Conference Paper
AggregationSummary of expirements exploring the performance of Bureaucratic and Democratic Meta-Learners.  Areas explored so far: Number of base learners, K-Values.  Other areas to explore are suggested at the end.
Icon
MachineLearningDistro4
GeneralA bug fix distribution containing source, data, javadocs, and experiment specification files.  First, DELETE WekaWeka.jar and RunExp.bat in your working directory. Download and unzip this file.  Then copy Distro4 'on-top-of' your directories.
Icon
MachineLearningDistro3
GeneralThe third distribution containing source, data, javadocs, and experiment specification files.  Download and unzip this file.  Then copy the contents 'on-top-of' the previous distribution's contents. This distro has been on the class web site for a week.
Icon
EXP1
AggregationRead the _README.TXT file.

Vary meta-learner test set % and master test set %

File includes experiement KOML and resulting CSV files.

Results summarized in an Excel file.

Further analysis to come later.
Icon
Project
AggregationThis is a short paper I wrote up to try to understand this project's scope.  Its pretty high level, but Dr. White asked me to post it for the rest of the class so here it is.
Icon
GenDoc-1.0-beta5
AggregationA quite nice xml editor.  Probably the best so far!
Icon
jaxe-windows-2.2
AggregationPerhaps significantly better than Pollo!  Let me know what you think.
Icon
pollo-0.4-bin
AggregationFree XML Editory
Icon
MachineLearningDistro2
General2nd Distribution of Weka (slightly extended), example experiment specification files, Batch files, data files, etc
New New Document
|
Upload Upload Document
|
Filter Filter
|  
TypeNameProjectDescriptionFilter
Icon
RandomForestsIntroduction
BriemanA very good introduction to Random Forests and the valuable data analysis that is possible.
Icon
randomforest2001
BriemanA paper by Brieman with additional information about random forests and the "random trees" within a random forests.
Icon
CaruanaEnsembleModels
AggregationCurana's paper on using 2000 mls.   There were only ~20 types of mls.  However, for each ml type, many different variations were created by setting the ml parameters to different values.
New New Item
|
Filter Filter
 
EndTitleDescriptionFilter
IMPORTANTExplain your ideas, methods, and answer others' questions.  We need your contributions.  You'll get significant credit if you do a good job.  NO NEGATIVE credit!!
IMPORTANT!!Ask questions -- get the BALL ROLLING -- Always! Be a leader!
New New Item
|
Filter Filter
|
Import from Address Book Import Contacts
 
Last NameFirst NameE-mail Address
Baldwin
Matthew
Basnight
Tomtbasnigh@harris.com
Booth
Taylortebooth@ncsu.edu
Burke
David
Gupta
Rachanarachana_a_gupta@yahoo.com
Haydar
Abdulateefahaydar@ncsu.edu
Hong
Taothong@ncsu.edu
Hungria
Andersonandyred49@hotmail.com
Livingston
Fredfjliving@ncsu.edu
McCauley
Evanetmccaul@ncsu.edu
Palmer
Ryan
Patel
Tejasthpatel@ncsu.edu
Pater
Nathannapater@ncsu.edu
Simmons
Jonjesimmo2@ncsu.edu
Wong
Jasonjawong@ncsu.edu
Wright
Timothytimothy.wright@gmail.com
 
 
A Human Learning Class!!
New New Item
|
Filter Filter
|
Change Order Change Order
 
  Class Project real-time publishing site (i.e., a Blog-Wiki-Weka)
  ECE 591Q / 492Q web site
  Mark White's Home Page
  Brieman's web site on Random Forests and Bagging
  Paper talking about combining bagging and boosting
New New Document
|
Upload Upload Document
|
New Folder New Folder
|
Filter Filter
|  
TypeNameLast Name
Icon
Thomas Basnight Journal Paper
Basnight
Icon
TaylorBooth_JournalPaper
Booth
Icon
journal_paper_Burke
Burke
Icon
ECE591Q_JournalPaper_Rachana
Gupta
Icon
Tao Journal paper V1
Hong
Icon
Anderson_Hungria_Conf_Paper
Hungria
Icon
JournalPaper_Livingston
Livingston
Icon
EvanJournalPaperFinal
McCauley
Icon
journal paper-rmpalmer
Palmer
Icon
Journal_Paper_Tejas_Patel
Patel
Icon
JournalPaper_Nathan_Pater
Pater
Icon
SimmonsJournalPaper
Simmons
Icon
Paper_JasonWong
Wong
Icon
Tim_Wright_Journal_Paper
Wright
New New Document
|
Upload Upload Document
|
New Folder New Folder
|
Filter Filter
|  
TypeNameLast Name
Icon
MattBaldwinJournalPaper
Baldwin
Icon
JournalFiles-rmpalmer
Baldwin
Icon
TAB Experiments and Results
Basnight
Icon
taylorBoothsExperimentPackage
Booth
Icon
JournalPaperFiles
Gupta
Icon
Tao Hong 1
Hong
Icon
Tao Hong 2
Hong
Icon
Tao workingDir
Hong
Icon
DataFiles
Hungria
Icon
Weka_Distro_FredLivingston
Livingston
Icon
RandomForest
Livingston
Icon
RAFT_Distro_Livingston
Livingston
Icon
EvanMcCauleyJournalPaperFiles
McCauley
Icon
FinalPaperFiles
McCauley
Icon
NPater-Weka_with_Proximities
Pater
Icon
SimmonsData
Simmons
Icon
jawong
Wong
Icon
Timothy_Wright_Journal_Paper_Files
Wright
Icon
Palmer
New New Document
|
Upload Upload Document
|
New Folder New Folder
|
Filter Filter
|  
TypeConference Paper's Author
IconAbdulateef HAYDAR
IconAbdulateef HAYDAR
IconAbdulateef HAYDAR
IconAbdulateef HAYDAR
IconAnderson HUNGRIA
IconAnderson HUNGRIA
IconAnderson HUNGRIA
IconAnderson HUNGRIA
IconAnderson HUNGRIA
IconDavid BURKE
IconDavid BURKE
IconDavid BURKE
IconFred LIVINGSTON
IconFred LIVINGSTON
IconJason WONG
IconJason WONG
IconMark White
IconMatthew BALDWIN
IconMatthew BALDWIN
IconNathan PATER
IconNathan PATER
IconNathan PATER
IconRachana GUPTA
IconRachana GUPTA
IconRachana GUPTA
IconRachana GUPTA
IconRachana GUPTA
IconRachana GUPTA
IconRachana GUPTA
IconRyan PALMER
IconRyan PALMER
IconTao HONG
IconTao HONG
IconTao HONG
IconTao HONG
IconTaylor BOOTH
IconTaylor BOOTH
IconTaylor BOOTH
IconTejas PATEL
IconTejas PATEL
IconTejas PATEL
IconTim WRIGHT
IconTim WRIGHT
IconTim WRIGHT
IconTim WRIGHT
IconTim WRIGHT
IconTim WRIGHT
IconTom BASNIGHT
IconTom BASNIGHT
IconTom BASNIGHT
New New Document
|
Upload Upload Document
|
New Folder New Folder
|
Filter Filter
|  
TypeNameGoal(s)
Created By
Icon
MLConfPaper
Aggregation
Matthew BALDWIN
Icon
Use_of_Filters_to_Improve_Accuracy_of_Machine_Learners
Aggregation
Tim WRIGHT
Icon
Use_of_Filters_to_Improve_Accuracy_of_Machine_Learners
Aggregation
Tim WRIGHT
Icon
ConferencePaper_Nathan_Pater
Brieman
Nathan PATER
Icon
EvanAndyAnalysisFiles
Aggregation
Evan MCCAULEY
Icon
EvanAndyConferencePaperV1
Aggregation
Evan MCCAULEY
Icon
haydar Files+paper
Aggregation
Abdulateef HAYDAR
Icon
SimmonsConferencePaper
Aggregation
Jonathan SIMMONS
Icon
ConferencePaper
Aggregation
Taylor BOOTH
Icon
Conference Paper
Aggregation
Abdulateef HAYDAR
Icon
ConferencePaper-Fred
Brieman
Fred LIVINGSTON
Icon
rmpalmer-conference
Aggregation
Ryan PALMER
Icon
David_Burke_and_Tim_Wright_Conf_Paper_Files
Aggregation
Tim WRIGHT
Icon
David_Burke_Tim_wright_conf_paper
Aggregation
Tim WRIGHT
Icon
Conference_Paper_Tejas_Patel
Aggregation
Tejas PATEL
Icon
MachineLearningDistroFred-alpha
Brieman
Fred LIVINGSTON
Icon
Tao Hong's Experiments
Aggregation
Tao HONG
Icon
Tao Hong_Conference paper V1
Aggregation
Tao HONG
Icon
Results_Tom_Basnight
Aggregation
Tom BASNIGHT
Icon
DataFiles_JasonWong
Aggregation
Jason WONG
Icon
ECE591q_ConferencePaper_Rachana
Aggregation
Rachana GUPTA
Icon
Aggregation_Project_Excel_komlandOtherFiles_Rachana
Aggregation
Rachana GUPTA
Icon
Paper_JasonWong
Aggregation
Jason WONG