Test Bed

   
Small Data Sets

A small data set can either be a graph (g), image collection (i), dissimilarity matrix (m), or a data table (t).

Table B.1: Small data collections

name	size	type	source
aud	18	m	Mietta E. Lennes, Department of Phonetics, University of Helsinki
bridges	108	t	UCI Machine Learning Repository (MLR) [Blake98]
crcars	406	t	D. Donoho [Donoho83]
cars	38	t	H. V. Henderson [Henderso81]
gd98c	62	g	Graph Drawing '98 Contest
corel-244	100	i	Kerry Rodden, Computer Laboratory, University of Cambridge
corel-385	100	i	Kerry Rodden
cpu-performance	209	t	MLR
dermatology	366	t	MLR
detroit	13	t	StatLib Datasets Archive
echocardiogram	131	t	MLR
ecoli	107	g	Karen Eilbeck, Biochemistry Division, University of Manchester
FFT	18	m	Mietta E. Lennes
flags	194	t	MLR
gd99c	105	g	Graph Drawing '99 Contest
gene2sc	112	t	MLR
glass	214	t	MLR
GPA1component	27	g	Karen Eilbeck
group	16	t	Opera Group, Computer Laboratory, University of Cambridge
haberman	306	t	MLR
heart	270	t	MLR
house-votes	435	t	MLR
housing	506	t	MLR
humandevel	130	t	MLR
image	210	t	MLR
imports-85	205	t	MLR
ionosphere	351	t	MLR
iris	150	t	MLR
Kellog	23	t	T. Cox [Cox94]
letters	26	t	F. Labelle's Dimensionality Reduction page
letters-back	24	t	F. Labelle
liver-disorders	345	t	MLR
misc	206	t	MLR
network	16	g	J. B. Kruskal [Kruskal78a], Figure 3: ``Input for Corn Biomass Network''
odmg-schema	22	g	The Object Database Standard version 2.0 [Cattell97]
o-ring-erosion	23	t	MLR
places	329	t	StatLib
planets	9	t	F. Labelle
post-operative	90	t	MLR
protein	25	t	Handbook of Small Data Sets [Hand94]
query	20	t	MLR
retention	270	t	U.S. Department of Agriculture
servo	167	t	MLR
shuttle	15	t	MLR
Skulls	40	t	T. Cox
solar-flare	323	t	MLR
soybean	307	t	MLR
tae	151	t	MLR
usa-sales	26	t	American Automobile Manufacturers' Association (AAMA)
usa-share	26	t	AAMA
wine	178	t	MLR
Yoghurt	12	t	T. Cox
zoo	101	t	MLR
minimum	9
maximum	506
mean	145
median	107

   
Medium Data Sets

All medium-sized data sets in the test bed are tabular, and come from the UCI Machine Learning Repository[Blake98].

Table B.2: Medium-sized data tables

name	size
australian	690
breast-cancer	699
cloud-1	1024
cloud-2	1024
cmc	1473
credit-screening	690
german	1000
pima-indians-diabetes	768
tic-tac-toe	958
vehicle	846
vowel	990
water-treatment	527
yeast	1484
minimum	527
maximum	1484
mean	936
median	958

   
Large Data Sets

All large data sets in the test bed are tabular. The number of attributes (columns) is also given, as it is relevant for Principal Components Analysis (see Section 4.3), which was applied to them.

Table B.3: Large data tables

name	rows	cols	source
abalone	4177	8	UCI Machine Learning Repository (MLR) [Blake98]
letter	20000	16	MLR
mfeat	2000	6	MLR
pageblock	5473	10	MLR
sat	6435	36	MLR
segment	2310	18	MLR
nbody-1	15000	12	Jarrod Hurley, Institute of Astronomy, University of Cambridge
nbody-2	14898	12	Jarrod Hurley
shuttle	58000	9	MLR
spambase	4601	57	MLR
minimum	2000
maximum	58000
mean	13289
median	5954