ocaml_gpr

ocaml_gpr

################################
# train
$ ocaml_gpr -verbose -cmd train -model foo.model < train.data
target variance: 53980.91115
iter    1: MSLL=20677.0739875 SMSE=0.6434589 MAD=144.2473817 MAXAD=395.0879824
iter    1: |gradient|=406648.85928
result: MSLL=0.0000006 SMSE=1.0000000 MAD=186.1172019 MAXAD=488.0434778

####################################################
# Now apply, all data is just the mean of the series
$ ocaml_gpr -cmd test -model foo.model < test.csv
1556.043478
1556.043477
1556.043477
1556.043477

#####
Training data
$ cat train.data
0.0,1767.0
1.0,1643.0
2.0,1605.0
3.0,1675.0
4.0,1717.0
5.0,1505.0
6.0,1746.0
7.0,1864.0
8.0,1619.0
9.0,1599.0
10.0,1593.0
11.0,1831.0
12.0,1825.0
13.0,1707.0
14.0,1562.0
15.0,1728.0
16.0,1433.0
17.0,1392.0
18.0,1194.0
19.0,1068.0
20.0,1482.0
21.0,1160.0
22.0,1074.0

#####
Test input - handful of out of sample data points
$ cat test.csv
23.0
24.0
25.0
26.0

###########
#####
### 'sta' is a little command line descriptive statistics app (https://github.com/simonccarter/sta)
$ cat train.data | cut -d',' -f2 | sta
N       min     max     sum     mean    sd      sderr
23      1068    1864    35789   1556.04 232.338 48.4458