[pymvpa] running pyMVPA in the virtual unix

Yaroslav Halchenko debian at onerussian.com
Wed Dec 17 22:02:33 UTC 2008


> >> What is 'extremely slow' in seconds (or minutes)? ;-) I've been running 
> >> Tiny example of full-cross validation for the last half an hour and am 
> >> still in the middle (dataset=datset.selectSamples( ....) So, I guess it 
ho ho -- half an  hour! wow...

try running

MVPA_DEBUG_METRICS=asctime,reltime,vmem MVPA_DEBUG=CROSSC,CERR,CLF,CLF_ doc/examples/start_easy.py

then you should see details like

[CLF   ] DBG{Wed Dec 17 17:00:28 2008 / 0.503 sec / VmSize: 266888 kB}:     Training classifier SMLR() / SMLR: 8 states:states{trained_dataset predicting_time+ training_confusion predictions+...} 11 params:params{has_bias=True fit_all_weights=True lm=0.1 min_resamp=0.001...} on dataset Dataset / float32 108 x 530 uniq: 6 chunks 2 labels
[CLF   ] DBG{Wed Dec 17 17:00:29 2008 / 1.001 sec / VmSize: 266888 kB}:     Predicting classifier SMLR() / SMLR: 8 states:states{trained_dataset predicting_time+ training_confusion predictions+...} 11 params:params{has_bias=True fit_all_weights=True lm=0.1 min_resamp=0.001...} on data (108, 530)
[CERR  ] DBG{Wed Dec 17 17:00:29 2008 / 0.006 sec / VmSize: 266888 kB}:   Classifier error on Dataset / float32 108 x 530 uniq: 6 chunks 2 labels: 0.06
[CROSSC] DBG{Wed Dec 17 17:00:29 2008 / 0.001 sec / VmSize: 266888 kB}:  Split #0: result 0.0555555555556
[CLF   ] DBG{Wed Dec 17 17:00:29 2008 / 0.005 sec / VmSize: 267336 kB}:     Training classifier SMLR() / SMLR: 8 states:states{trained_dataset predicting_time*+ training_confusion predictions*+...} 11 params:params{has_bias=True fit_all_weights=True lm=0.1 min_resamp=0.001...} on dataset Dataset / float32 108 x 530 uniq: 6 chunks 2 labels
[CLF   ] DBG{Wed Dec 17 17:00:30 2008 / 0.876 sec / VmSize: 267336 kB}:     Predicting classifier SMLR() / SMLR: 8 states:states{trained_dataset predicting_time+ training_confusion predictions+...} 11 params:params{has_bias=True fit_all_weights=True lm=0.1 min_resamp=0.001...} on data (108, 530)
[CERR  ] DBG{Wed Dec 17 17:00:30 2008 / 0.006 sec / VmSize: 267336 kB}:   Classifier error on Dataset / float32 108 x 530 uniq: 6 chunks 2 labels: 0.06
[CROSSC] DBG{Wed Dec 17 17:00:30 2008 / 0.001 sec / VmSize: 266888 kB}:  Split #1: result 0.0555555555556

So as you can see -- here most of the time was spent on training the classifier
(reltime shows difference in time to previous printed line): 1.001 sec and
0.876 sec.

I wonder what numbers are you getting, or where does it get stuck

-- 
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student  Ph.D. @ CS Dept. NJIT
Office: (973) 353-1412 | FWD: 82823 | Fax: (973) 353-1171
        101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW:     http://www.linkedin.com/in/yarik        



More information about the Pkg-ExpPsy-PyMVPA mailing list