I mean

• type of a variable (float, int or categorical)
• its distribution

@tsuyos-u
The tutorial (examples/grain_bound/tutorial.ipynb) tries to download data file (s5-210.csv) from www.tudalab.org. However, currently access to the file is rejected by the web server:

--2019-11-27 17:15:03--  https://tsudalab-public-files.s3-ap-northeast-1.amazonaws.com/s5-210.csv
Resolving tsudalab-public-files.s3-ap-northeast-1.amazonaws.com (tsudalab-public-files.s3-ap-northeast-1.amazonaws.com)... 52.219.68.139
Connecting to tsudalab-public-files.s3-ap-northeast-1.amazonaws.com (tsudalab-public-files.s3-ap-northeast-1.amazonaws.com)|52.219.68.139|:443... connected.
HTTP request sent, awaiting response... 403 Forbidden
2019-11-27 17:15:03 ERROR 403: Forbidden.

Could you please update the URL of the data file, if it's place has been changed?

Hello developers,

I have a question about extraction of fmean and fcov after learning.

fmean = policy.predictor.get_post_fmean(policy.training, policy.test)
fcov = policy.predictor.get_post_fcov(policy.training, policy.test)

The above scripts worked when num_rand_basis was not set at 0.
However, it did not work when num_rand_basis was set at 0.

How can I get fmean and fcov values when GP (num_rand_basis = 0) was conducted?

The error message was as below

ValueError Traceback (most recent call last)
in ()
----> 1 fmean = policy.predictor.get_post_fmean(policy.training, policy.test)
2 fcov = policy.predictor.get_post_fcov(policy.training, policy.test)

/usr/local/lib/python2.7/dist-packages/combo/gp/predictor.pyc in get_post_fmean(self, training, test)
32 if self.model.stats is None:
33 self.prepare( training )
---> 34 return self.model.get_post_fmean( training.X, test.X )
35
36 def get_post_fcov( self, training, test, diag = True ):

/usr/local/lib/python2.7/dist-packages/combo/gp/core/model.pyc in get_post_fmean(self, X, Z, params)
102
103 if self.inf is 'exact':
--> 104 post_fmu = inf.exact.get_post_fmean(self, X, Z, params)
105
106 return post_fmu

/usr/local/lib/python2.7/dist-packages/combo/gp/inf/exact.pyc in get_post_fmean(gp, X, Z, params)
92 G = gp.prior.get_cov( X=Z, Z=X, params = prior_params )
93
---> 94 return G.dot(alpha) + fmu
95
96 def get_post_fcov(gp, X, Z, params = None, diag = True ):

ValueError: shapes (81,22) and (21,) not aligned: 22 (dim 1) != 21 (dim 0)

X = combo.misc.centering( X )

whenever I am trying to run the above line in spyder, it's showing one error as: 'AttributeError: module 'combo' has no attribute 'misc''.

Can anyone please give a solution of this!
Thanks in advance.

I have already installed numpy >=1.10, scipy >= 0.16 and Cython >= 0.22.1. But still get the ModuleNotFoundError.
Could anyone kindly tell me is there any dependency for the ‘gp’ here？

This version of combo does not currently have the following cythonized files to run COMBO.
in combo/combo/gp/_src:

enhance_gauss.so

in combo/combo/misc/_src:

cholupdate.so
diagAB.so
logsumexp.so
traceAB.so

These files can be easily generated via cython but the current version of combo will throw an ImportError because it cannot find them, as only the .pyx and .c files are currently included.

max_num_probes -> num_rounds
num_search_each_probe -> num_probes

Hi developers of combo,

Thanks for providing a great library. I want to use it for my projects.

I have a question about multiple evaluations at each probe. For a function f(x) which takes a long time to run, I want to compute f(x) for a number of x values in parallel. To test this, I ran tutorial.ipynb with the following modifications:

res = policy.random_search(max_num_probes=5, simulator=simulator(), 
                           num_search_each_probe=4)

# Originally, max_num_probes=80 and num_search_each_prob=1
# I expect that the total amount of computation is not changed
res = policy.bayes_search(max_num_probes=20, simulator=simulator(), score='TS', 
                          interval=5, num_rand_basis=5000, num_search_each_probe=4)

Is this the right way to do parallel computation as explained above?

And I'm concerned about actual computation time. The modified code seems slower
than the original, specifically on policy.bayes_search():

Original: 99.50355625152588 [sec]
Modified: 1856.6796779632568 [sec]

I expected the computation time is proportional to the number of evaluations of f(x),
so I don't understand the significant difference. That's why I'm wondering that I'm not correct
on using parallel evaluation.

Everything is installed, but the files in the run case are still not working. Has anyone solved the problem yet?

File "/home/wjw/python-file/666.py", line 4, in
import combo
File "/home/wjw/anaconda3/envs/COMBO1/lib/python2.7/site-packages/combo/init.py", line 1, in
import gp
File "combo/gp/cov/_src/enhance_gauss.pyx", line 1, in init combo.gp.cov._src.enhance_gauss
ValueError: numpy.ufunc size changed, may indicate binary incompatibility. Expected 216 from C header, got 192 from PyObject