timbl(1) | General Commands Manual | timbl(1) |
NAME¶
timbl - Tilburg Memory Based Learner
SYNOPSIS¶
timbl [options]
timbl -f data-file -t test‐file
DESCRIPTION¶
TiMBL is an open source software package implementing several memory‐based learning algorithms, among which IB1‐IG, an implementation of k‐nearest neighbor classification with feature weighting suitable for symbolic feature spaces, and IGTree, a decision‐tree approximation of IB1‐IG. All implemented algorithms have in common that they store some representation of the training set explicitly in memory. During testing, new cases are classified by extrapolation from the most similar stored cases.
OPTIONS¶
-a <n> or -a <string>
Possible values are:
0 or IB
the IB1 (k‐NN) algorithm (default)
1 or IGTREE
a decision‐tree‐based approximation of IB1
2 or TRIBL
a hybrid of IB1 and IGTREE
3 or IB2
an incremental editing version of IB1
4 or TRIBL2
a non‐parameteric version of TRIBL
-b n
-B n
--Beam=<n>
--clones=<n>
-c n
+D
--Diversify
-d val
Z : equal weights to all (default)
ID : Inverse Distance
IL : Inverse Linear
ED:a : Exponential Decay with factor a (no whitespace!)
ED:a:b : Exponential Decay with factor a and b (no whitespace!)
-e n
-f file
-F format
-G normalization
Supported normalizations are:
Probability or 0
normalize between 0 and 1
addFactor:<f> or 1:<f>
add f to all possible targets, then normalize between 0 and 1 (default f=1.0).
logProbability or 2
Add 1 to the target Weight, take the 10Log and then normalize between 0 and 1
+H or -H
-i file
-I file
-k n
-L n
-l n
-m string
The format is : GlobalMetric:MetricRange:MetricRange
e.g.: mO:N3:I2,5-7
C: cosine distance. (Global only. numeric features implied)
D: dot product. (Global only. numeric features implied)
DC: Dice coefficient
O: weighted overlap (default)
E: Euclidian distance
L: Levenshtein distance
M: modified value difference
J: Jeffrey divergence
S: Jensen‐Shannon divergence
N: numeric values
I: Ignore named values
--matrixin=file
--matrixout=file
-n file
-M n
-N n
-o s
--occurrences=<value>
-O path
-p n
-P path
-q n
-R n
-s
-s0
-T n
-t file
-t leave_one_out
-t cross_validate
-t @file
--Treeorder =value n
DO: none
GRO: using GainRatio
IGO: using InformationGain
1/V: using 1/# of Values
G/V: using GainRatio/# of Valuess
I/V: using InfoGain/# of Valuess
X2O: using X‐square
X/V: using X‐square/# of Values
SVO: using Shared Variance
S/V: using Shared Variance/# of Values
GxE: using GainRatio * SplitInfo
IxE: using InformationGain * SplitInfo
1/S: using 1/SplitInfo
-u file
-U file
-V
+v level or -v level
s: work silently
o: show all options set
b: show node/branch count and branching factor
f: show calculated feature weights (default)
p: show value difference matrices
e: show exact matches
as: show advanced statistics (memory consuming)
cm: show confusion matrix (implies +vas)
cs: show per‐class statistics (implies +vas)
cf: add confidence to output file (needs -G)
di: add distance to output file
db: add distribution of best matched to output file
md: add matching depth to output file.
k: add a summary for all k neigbors to output file (sets -x)
n: add nearest neigbors to output file (sets -x)
You may combine levels using '+' e.g. +v p+db or -v o+di
-w n
0 or nw: no weighting
1 or gr: weigh using gain ratio (default)
2 or ig: weigh using information gain
3 or x2: weigh using the chi‐square statistic
4 or sv: weigh using the shared variance statistic
5 or sd: weigh using standard deviation. (all features must be numeric)
-w file
-w file:n
-W file
+% or -%
+x or -x
(IB1 and IB2 only, default is -x)
-X file
BUGS¶
possibly
AUTHORS¶
Ko van der Sloot Timbl@uvt.nl
Antal van den Bosch Timbl@uvt.nl
SEE ALSO¶
2017 November 9 |