diff options
| author | Benjamin Franzke <benjaminfranzke@googlemail.com> | 2012-05-10 14:19:12 +0200 |
|---|---|---|
| committer | Benjamin Franzke <benjaminfranzke@googlemail.com> | 2012-05-10 14:22:40 +0200 |
| commit | 684e283468e768701b422a2bff4248c6c86b11f2 (patch) | |
| tree | 94e8dd7c456fac84ad9ace31a7eb5fb2c5820893 /run_tests.m | |
| parent | cdc7ef4b0a7b4581072771540b7807a03879346a (diff) | |
| download | dmc-684e283468e768701b422a2bff4248c6c86b11f2.tar.gz dmc-684e283468e768701b422a2bff4248c6c86b11f2.tar.bz2 dmc-684e283468e768701b422a2bff4248c6c86b11f2.zip | |
Combine logarithmic and quad prediction
Product 335 scales better with quadratic prediction.
Diffstat (limited to 'run_tests.m')
| -rw-r--r-- | run_tests.m | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/run_tests.m b/run_tests.m index 065e1e2..c88ea60 100644 --- a/run_tests.m +++ b/run_tests.m @@ -19,6 +19,7 @@ random_data = rand_pred(train_data, 14); % quadratic just for reference, it sucks more than mean-predicition [quqerr, quterr] = calc_error('quad reg.',real_data, quad_data); [loqerr, loterr] = calc_error('log reg.', real_data, log_data); +[loqerr, loterr] = calc_error('q+l reg.', real_data, [ log_data(:,1:334) quad_data(:,335) log_data(:,336:570) ]); [seqerr, seterr] = calc_error('sevenday', real_data, sevenday_data); [raqerr, raterr] = calc_error('random', real_data, random_data); |