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
|
---------------------------------------------------
Some examples to learn how to *program* in Torch3.
Of course you can use them without any changes,
but you will not use 1/10 of the library if you
don't want to code something!
boosting.cc: adaboost with MLP in classification.
knn.cc: K-nearest-neighbors algorithm.
mlp.cc: multi-layered perceptron.
svm.cc: support vector machines.
svm_multi.cc: SVM with more than two classes.
---------------------------------------------------
You will find two examples of datasets in data/.
Torch3 is able to deal directly with gziped data,
but you can uncompress them if you want to have a
look. (Or do a zless, zcat, zmore).
These data are classification data: there is 784
inputs (the first 784 columns) and the 1 target
which can take values between 0..9. (10 classes).
---------------------------------------------------
Examples of command line:
Linux_OPT_FLOAT/boosting -nhu 5 -iter 20 -n 5 -valid data/test_data.gz data/train_data.gz 784 10
Linux_OPT_FLOAT/knn -kfold 10 -K 3 data/train_data.gz 784 10
Linux_OPT_FLOAT/mlp -nhu 10 -iter 20 -save model data/train_data.gz 784 10
Linux_OPT_FLOAT/mlp --test model data/test_data.gz
Linux_OPT_FLOAT/mlp -kfold 10 -nhu 10 -iter 20 data/train_data.gz 784 10
Linux_OPT_FLOAT/svm -std 1650 -class 9 data/train_data.gz model
Linux_OPT_FLOAT/svm --test model data/test_data.gz
Linux_OPT_FLOAT/svm --kfold -std 1650 -class 9 data/train_data.gz 10
Linux_OPT_FLOAT/svm_multi -std 1650 data/train_data.gz model 10
Linux_OPT_FLOAT/svm_multi --test model data/test_data.gz
For each command line, results are in the_class_err or the_valid_class_err.
---------------------------------------------------
|
