yogsototh/clj-ml
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Documentation for clusterers and small fixes in the clusterer options for :expectation-maximization clusterer

Browse source
This commit is contained in:
Antonio Garrote 2010-03-16 09:22:49 +01:00
parent 06adde17e9
commit 47e78f5fb4
4 changed files with 115 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
src/clj_ml/classifiers.clj
View file

@ -156,7 +156,7 @@
The first argument identifies the kind of classifier and the second
argument the algorithm to use, e.g. :decission-tree :c45.
The colection of classifiers currently supported are:
The classifiers currently supported are:
- :decission-tree :c45
- :bayes :naive
@ -459,7 +459,23 @@
The function returns the newly classified instance.
This call is destructive, the instance passed as an argument
is modified."
is modified.
; We create the instance to classify
(def *to-classify* (make-instance *dataset* {:class :Iris-versicolor
:petalwidth 0.2
:petallength 1.4
:sepalwidth 3.5
:sepallength 5.1}))
; We use the classifier to check the value for the class
(classifier-classify *classifier* *to-classify*)
>0.0
; We change the class for the instance according to the assigned class
(classifier-label *classifier* *to-classify*)
>#<Instance 5.1,3.5,1.4,0.2,Iris-setosa>
"
([classifier instance]
(let [cls (classifier-classify classifier instance)]
(instance-set-class instance cls))))

78
src/clj_ml/clusterers.clj
View file

@ -7,11 +7,12 @@
clj-ml.clusterers
"This namespace contains several functions for
building clusterers using different clustering algorithms. K-means, Cobweb and
Expectation maximization algorithms are currently supported.
Expectation maximization algorithms are currently supported. Some of these
algorithms support incremental building of the clustering without having the
full data set in main memory. Functions for evaluating the clusterer as well
as for clustering new instances are also supported"
Some of these algorithms support incremental building of the clustering without
having the full data set in main memory. Functions for evaluating the clusterer
as well as for clustering new instances are also supported
"
(:use [clj-ml utils data distance-functions]
[incanter charts])
(:import (java.util Date Random)
@ -20,7 +21,8 @@
;; Setting up clusterer options
(defmulti make-clusterer-options
(defmulti #^{:skip-wiki true}
make-clusterer-options
"Creates ther right parameters for a clusterer"
(fn [kind map] kind))
@ -51,8 +53,9 @@
(defmethod make-clusterer-options :expectation-maximization
([kind map]
(let [cols-val-a (check-option-values {:acuity "-A"
:cutoff "-C"
(let [cols-val-a (check-option-values {:number-clusters "-N"
:maximum-iterations "-I"
:minimum-standard-deviation "-M"
:random-seed "-S"}
map
[""])]
@ -78,7 +81,66 @@
clusterer#)))
(defmulti make-clusterer
"Creates a new clusterer for the given kind algorithm and options"
"Creates a new clusterer for the given kind algorithm and options.
The first argument identifies the kind of clusterer. The second argument
is a map of parameters particular to each clusterer.
The clusterers currently supported are:
- :k-means
- :cobweb
- :expectation-maximization
This is the description of the supported clusterers and the parameters accepted
by each clusterer algorithm:
* :k-means
A clusterer that uses the simple K-Means algorithm to build the clusters
Parameters:
- :display-standard-deviation
Display the standard deviation of the centroids in the output for the
clusterer. Sample value: true
- :replace-missing-values
Replaces the missing values with the mean/mode. Sample value: true
- :number-clusters
The number of clusters to be built. Sample value: 3
- :random-seed
Seed for the random number generator. Sample value: 0.3
- :number-iterations
Maximum number of iterations that the algorithm will run. Sample value: 1000
* :cobweb
Implementation of the Cobweb incremental algorithm for herarchical conceptual clustering.
Parameters:
- :acuity
Acuity. Default value: 1.0
- :cutoff
Cutoff. Default value: 0.002
- :random-seed
Seed for the random number generator. Default value: 42.
* :expectation-maximization
Implementation of the probabilistic clusterer algorithm for expectation maximization.
Parameters:
- :number-clusters
Number of clusters to be built. If ommitted or -1 is passed as a value, cross-validation
will be used to select the number of clusters. Sample value: 3
- :maximum-iterations
Maximum number of iterations the algorithm will run. Default value: 100
- :minimum-standard-deviation
Minimum allowable standard deviation for normal density computation. Default value: 1e-6
- :random-seed
Seed for the random number generator. Default value: 100
"
(fn [kind & options] kind))

7
src/clj_ml/data_store.clj
View file

@ -29,7 +29,12 @@
format))))
(defmulti make-data-store-connection
"Connects to a data store"
"Connects to a data store.
- The first parameter is the kind of data store to connect to.
- The second parameter is a map with options for the connection
to that kind of data store.
"
(fn [kind params] kind))
(defmethod make-data-store-connection :mongodb

21
test/clj_ml/clusterers_test.clj
View file

@ -26,6 +26,27 @@
(is (= (aget options 9)
"1"))))
(deftest make-clusterers-options-expectation-maximization
(let [options (make-clusterer-options :expectation-maximization {:number-clusters 3 :maximum-iterations 10 :minimum-standard-deviation 0.001 :random-seed 30})]
(is (= (aget options 0)
""))
(is (= (aget options 1)
"-N"))
(is (= (aget options 2)
"3"))
(is (= (aget options 3)
"-I"))
(is (= (aget options 4)
"10"))
(is (= (aget options 5)
"-M"))
(is (= (aget options 6)
"0.0010"))
(is (= (aget options 7)
"-S"))
(is (= (aget options 8)
"30"))))
(deftest make-and-build-clusterer
(let [ds (make-dataset :test [:a :b] [[1 2] [3 4]])