-
Notifications
You must be signed in to change notification settings - Fork 0
/
GridSearch - README.txt
71 lines (49 loc) · 3.22 KB
/
GridSearch - README.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
GRID-SEARCH.JAR
##########################################################
grid-search.jar searches for the best parameters that maximize the performance of a machine learning algorithm.
The program will output a .txt file with the performances found in each fold for each parameter searched.
The parameters that are found to lead to the best performances are used and a model.bin is returned. This is the best model
learned by the machine learning algorithm.
The grid-search.jar executable performs the following algorithms:
(based on Support Vector Machines)
SVMrank with Linear Kernel - Contains 1 parameter: the c parameter
SVMrank with Non-Linear Kernel - Contains 2 parameters: the c parameter and the g parameter
(based on Neural Networks)
RankNet - Contains 2 parameters: the number of iterations and the number of nodes to be present in the hidden layer
(based on Boosting Theory)
AdaRank - Contains 1 parameter: the number of iterations
RankBoost - Contains 2 parameters: the number of iterations and the threshold for weak candidates
(based on Constraints)
Coordinate Ascent - Contains 2 parameters: the number of random restarts and the number of iterations
General example of usage:
java -jar grid-search.jar SVMrankLinear <path> <c_param>
java -jar grid-search.jar SVMrankNonLinear <path> <c_param> <g_param>
java -jar grid-search.jar RankNet <path> <epochs_param> <nodes_param>
java -jar grid-search.jar AdaRank <path> <iterations_param>
java -jar grid-search.jar RankBoost <path> <iterations_param> <threshold_candidates_param>
java -jar grid-search.jar Coordinate_Ascent <path> <randomRestarts_param> <iterations_param>
To run the examples that I sent you, do the following.
RUN SVMRANK LINEAR KERNEL: train a model that searches for the best parameters until a maximum of 5 in the c parameter
######################################################################
cd examples
java -jar grid-search.jar SVMrankLinear ./experts/ 5
RUN SVMRANK NON LINEAR KERNEL: train a model that searches for the best parameters until a maximum of 2 in the c parameter and 2 in the g parameter
######################################################################
cd examples
java -jar grid-search.jar SVMrankNonLinear ./experts/ 2 2
RUN RANKNET: train a model that searches for the best parameters until a maximum of 10 iterations and at most 2 nodes in the hidden layer
######################################################################
cd examples
java -jar grid-search.jar RankNet ./experts/ 10 2
RUN ADARANK: train a model that searches for the best parameters until a maximum of 5 iterations
######################################################################
cd examples
java -jar grid-search.jar AdaRank ./experts/ 5
RUN RANKBOOST: train a model that searches for the best parameters until a maximum of 5 iterations and at most 2 weak candidates
######################################################################
cd examples
java -jar grid-search.jar RankBoost ./experts/ 5 2
RUN COORDINATE ASCENT: train a model that searches for the best parameters until 5 random restarts and 3 iterations
######################################################################
cd examples
java -jar grid-search.jar Coordinate_Ascent ./experts/ 5 3