professor

view ChangeLog @ 508:a27bd36750d2

Tiny code cleanup and some more documentation
author Holger Schulz <holger.schulz@durham.ac.uk>
date Thu, 07 Sep 2017 15:11:43 +0100
parents bde41e212fdb
children
line source
1 2017-08-22 Holger Schulz <holger.schulz@durham.ac.uk>
3 * Improvements to Ipol --- we allow different orders for each bin now,
4 there is also --order auto which does some jackknifing to figure out
5 the best suited polynomial order to avoid overfitting. The Ipol
6 outputformat is now binned 3 which dumps the minparam values and
7 maxparamvalues at the end of the coefficient string. This is necessary
8 as we now also allow to filter the inputs by median error (necessary
9 to tame NLO spiky histograms). The ranges are read back for each bin
10 separately now, happens in the C++ part.
12 * Overhaul of sampling, including things such as full support of sobol
13 sampling. Also, param range files with more than two numbers on a line
14 are interpreted as sub spaces for patch sampling.
16 * Add script prof2-ipolenvelopes which for each bin finds the minimum
17 and maximum polynomial value (using migrad) in the range of the
18 ipols validity. This is nice especially to check envelopes once
19 filtering in the ipol step was done
21 2017-05-18 Holger Schulz <holger.schulz@durham.ac.uk>
23 * Refitting of ipols using smeared ipol output.
24 Required some modification in data io structure.
26 2017-05-04 Holger Schulz <holger.schulz@durham.ac.uk>
28 * Fix in prof2-tune when checking analysis names, at the same time
29 allow for for closure test tuning, i.e. tune against input folder.
32 2017-05-02 Holger Schulz <holger.schulz@durham.ac.uk>
34 * Use -march=native according to eigen3 faq for better performance
35 as well as -O3 in pyext
37 * Replace modulus based status message by simple counter in ipol stage
39 * Set version to 2.2.2beta for imminent release
41 2017-04-28 Holger Schulz <holger.schulz@durham.ac.uk>
43 * Use BDSSVD in case Eigen >=3.3.3
45 * Reworked prof2-ipol and ipol.py to use mk_ipolbin rather than
46 mk_ipolhisto, some speedups plus additional benefit of being
47 able to fiddle with inividual bins in the ipol building stage
49 * add --minos switch to prof2-tune toi run MINOS
51 * Write GOF and NDOF into result file
53 2017-02-21 Andy Buckley <andy.buckley@cern.ch>
55 * Change prof2-ipol ierr mode default to 'symm'
57 2017-02-17 Andy Buckley <andy.buckley@cern.ch>
59 * Add a --limit-errs argument for prof2-tune, to avoid annoying
60 re-reading of large run dirs
62 * Tweaks to default error ipol order, and release figure handles
63 in envelope plotting.
65 2017-02-02 Holger Schulz <holger.schulz@durham.ac.uk>
67 * Update Copyright string
69 * Instead of exiting when prof2-ipol encounters nan coefficients in
70 value or error ipols, spit out a warning and ignore that histogram
71 entirely --- this makes prof2-ipol and friends much more robust
73 * tag as 2.2.1 and release.
75 2017-02-01 Holger Schulz <holger.schulz@durham.ac.uk>
77 * Use Agg as default matplotlib backend.
79 * Bump version to 2.2.1
81 2016-12-19 Holger Schulz <holger.schulz@cern.ch>
83 * Even more robust prof-tune, added --filter CL option to
84 automatically get rid of zero error bins
86 * Fixed a bug where IpolBin.n was never set which is needed for
87 the PointMatcher (i.e. weights syntax)
89 * Not using YODA Pointmatcher but own due to different regexes
90 required.
92 * Make release 2.2
94 2016-12-15 Holger Schulz <holger.schulz@cern.ch>
96 * More robust scripts and more meaningful error messages, especially
97 prof2-tune
99 2016-11-30 Holger Schulz <holger.schulz@cern.ch>
101 * Replace Counter code when running mk_Structure with much faster code
103 * Bump version to 2.2
105 * Bump DataFormat to binned 2
107 * Allow passing file names or open files (or similar streams) to
108 ipolio functions
110 * prof2-ipol can read yaml now
112 * Add some of Elliot's code for choosing sub spaces in prof2-ipol and
113 some plot routines to contrib
115 * Add SOBOL sampling to contrib
117 2016-08-26 Holger Schulz <holger.schulz@cern.ch>
119 * Use more readable Cython DL path in bootstrap
121 2016-07-18 Andy Buckley <andy.buckley@cern.ch>
123 * Typo fixes -- thanks to Leif Gellersen.
125 2016-07-14 Andy Buckley <andy.buckley@cern.ch>
127 * Release version 2.1.4
129 * Improve Makefile tests for ROOT and Cython versions.
131 2016-07-11 Andy Buckley <andy.buckley@cern.ch>
133 * Add Ipol min/max param vals getters and map them into Python.
135 * Add a built-in /REF-stripping option to data histo loading.
137 * Improvements to the Python ipol metadata interface.
139 2016-07-05 Andy Buckley <andy.buckley@cern.ch>
141 * Try to optimise long-vector building by not executing pow() if the exponent would be zero.
143 * Split C++ calcValue function into two, to allow re-use of a precalculated long-vector.
145 2016-06-23 Andy Buckley <andy.buckley@cern.ch>
147 * Split 'dot product' part of ipol value calculation into a standalone calcValue function.
149 * Use const references when passing const vectors in C++ API.
151 2016-04-14 Holger Schulz <holger.schulz@cern.ch>
153 * Introduce 1D sensitivity based on gradient
155 * User needs to specify --cmap or --grad when running prof2-sens now
157 2016-04-01 Holger Schulz <holger.schulz@cern.ch>
159 * Add a first version of the parametrisation using Machine Learning
161 2016-02-16 Holger Schulz <holger.schulz@cern.ch>
163 * Bump to version 2.1.3
165 2016-02-15 Holger Schulz <holger.schulz@cern.ch>
167 * prof2-sens: bugfix, add weight file usage, clean up CL
169 * prof2-I: add weight file usage
171 * prof2-predict: add weight file usage, fix output writing
173 * Extended documentation
175 2016-02-12 Holger Schulz <holger.schulz@cern.ch>
177 * Documentation including asciinema
179 2016-02-11 Holger Schulz <holger.schulz@cern.ch>
181 * Remove min_runs function, replace with numCoeffs
182 from C++ code
184 * Add prof2-envelopes script
186 * In prof-tune simplechi2 --- continue on weight==0
187 to allow some robustness when switching off bins
190 2016-02-10 Holger Schulz <holger.schulz@cern.ch>
192 * First version of prof2-runcombs plus sampling functions
193 salvaged from Prof1, modified to have no scipy dependence
195 * Add --rc switch to prof2-ipol to optionally read runcombs
196 Syntax is --rc runcoms.dat[colon]5
197 (Minimally invasive addon)
199 * Code cull in prof2-ipol. Only support multiprocessing
200 version, default 1 thread. Changed CL switch to '-j'.
203 2015-12-16 Holger Schulz <holger.schulz@cern.ch>
205 * Only calculate polynomials structure once per instance
206 -> significant speed increase
208 * Add the brute force grid scan (--scan) to prof2-tune
211 2015-12-11 Holger Schulz <holger.schulz@cern.ch>
213 * Minimiser tweaks --- no more need for param translation,
214 make strategy steerable (--strategy or -s), -q suppresses
215 iminuit output now, calculation and writing of correlation
216 matrix into results, dump weights into results file
218 2015-12-10 Holger Schulz <holger.schulz@cern.ch>
220 * Undo ui madness
222 * Remove numpy dependence (except sampling.py)
224 * Add prof-sens for sensitivity plotting exploting derivatives
226 2015-12-09 Holger Schulz <holger.schulz@cern.ch>
228 * Can specify PREFIX when calling make
230 2015-12-09 Andy Buckley <andy.buckley@cern.ch>
232 * Improve prof2-predict UI.
234 * Add metadata printing in prof2-lsipol debug mode.
236 * Add -v and -q flag shortcuts to all scripts.
238 2015-12-07 Holger Schulz <holger.schulz@cern.ch>
240 * Use __all__ in __init__.py
242 * Started sphinx documentation
244 * Support only iminuit, bump version to 2.2.alpha, restructure imports
246 * Sanity checks when calling parametrisation
248 2015-12-03 Andy Buckley <andy.buckley@cern.ch>
250 * Provide a version() function in the C++ library, and map it into
251 Python. Both sources use the VERSION variable defined in the
252 Makefile at build time.
254 * Reduce SVD fitting threshold to 1e-20.
256 2015-12-03 Holger Schulz <holger.schulz@cern.ch>
258 * Make derivative calculation work also when scaling params
260 * Let YODA write YODA
262 2015-12-02 Holger Schulz <holger.schulz@cern.ch>
264 * Derivative calculation
266 2015-11-28 Andy Buckley <andy.buckley@cern.ch>
268 * Version 2.1.2
270 * Revert range shift in Ipol.cc, since it produced (small)
271 numerical errors rather than improved stability. To be
272 understood...
274 * Add test/mkpolydata script, for generating polynomial pseudodata
275 used in closure tests.
277 2015-11-27 Andy Buckley <andy.buckley@cern.ch>
279 * Fix bug in histo loading.
281 * Adding log binning and other options to prof2-residuals.
283 2015-11-24 Andy Buckley <andy.buckley@cern.ch>
285 * Version 2.1.1 release.
287 * Fix prof2-predict YODA writing, and improve DataHisto.toYODA.
289 * Change parameter rescaling from using the [0,1] range to instead
290 use [1,2], to avoid arbitrarily scaled param numbers. BREAKS IPOL
291 FILE COMPATIBILITY WITH 2.1.0!!
293 * Set minimizer initial step size = prange/10.
295 * Add automatic determination of maximum MC errors in prof-tune,
296 and use them to regularize interpolated errs which could distort
297 the fit.
299 * Separate internal Minuit function names (with syntax
300 restrictions) from the free-form external param names.
302 2015-11-23 Andy Buckley <andy.buckley@cern.ch>
304 * Add pyext/professor2/chi2.py
306 * Add find_maxerrs() to the Python interface, for use in error regularisation.
308 * Use the param file name as a regex.
310 * Use __slots__ in the internal histogram and bin types, for better memory efficiency.
312 * Fix (re-fix?) use of inappropriate indices in calculation of mean and median MC uncertainties.
314 * More proactive data object deletion after conversion to YODA data format.
316 2015-11-22 Andy Buckley <andy.buckley@cern.ch>
318 * Version 2.1.0 (significant version change due to scaling, all-orders, and ROOT support)
320 2015-11-18 Andy Buckley <andy.buckley@cern.ch>
322 * dataio.py: Add ROOT file reading ability.
324 * Change rescaling I/O behaviour a bit, moving some hacked
325 Python-only Ipol string constructors into explicit API improvements, and
326 only activating rescaling if DoParamScaling is set true.
328 2015-11-16 Andy Buckley <andy.buckley@cern.ch>
330 * Add CPPFLAGS steering in Makefile and API tidying.
332 2015-11-11 Holger Schulz <holger.schulz@cern.ch>
334 * Add automatic raw param value -> unit range mapping to Ipol.
336 * Add any-order long vector calculation.
338 2015-11-10 Andy Buckley <andy.buckley@cern.ch>
340 * Add prof2-lsipol as a handy way to make a tuning weights file
341 and for general convenience.
343 2015-10-06 Andy Buckley <andy.buckley@cern.ch>
345 * Version 2.0.0
347 * Add 'make dist' target to Makefile, and tweak/reduce Cython necessity
349 * Allow calls to IpolBin and IpolHisto methods with params passed
350 as unpacked *args as well as a packed iterable.
352 * Add possibility to pass params as an dict-like as well as a list
353 or tuple of floats -- note that the dict-like must return
354 *ordered* values consistent with the Ipol training. Also ensure
355 the float type.
357 * Add conversion of C++ exceptions to Python exceptions.
359 * Add professor2.utils module, providing an opportunistic use of
360 OrderedDict for params so the params order required by Ipol is the
361 same as the order in which they are specified in the (first run's)
362 params.dat file.
364 * Change Python-mapped Ipol methods with no args to be properties.
366 * Add nice __repr__ methods for Histo and Bin objects.
368 * Specialise Histo as IpolHisto and DataHisto, and provide
369 toData{Histo,Bin} on IpolHisto and IpolBin respectively.
371 * Remove ProfMaster.
373 2015-10-04 Andy Buckley <andy.buckley@cern.ch>
375 * Add protection of svd.setThreshold call to ensure it is
376 supported/needed, via Eigen version number macros.
378 * Add vmin and vmax args to Ipol and IpolBin in Python.
380 2015-10-02 Andy Buckley <andy.buckley@cern.ch>
382 * Put histo file reading in a try..except block so read failures
383 on non-histo files in the run dirs are not fatal.
385 * Add protection in the histo loader, so we don't trip up on
386 Rivet's new cross-section and counter objects.
388 2015-09-30 Andy Buckley <andy.buckley@cern.ch>
390 * src/Ipol.h: Use Eigen/SVD include path rather than non-standard eigen3/Eigen/SVD.
392 2015-09-14 Holger Schulz <holger.schulz@durham.ac.uk>
394 * Use Eigen's svd.setThreshold(1e-20) to prevent the errors we saw with Eigen
395 3.2.2 and later when looking at 5th order polynomials in 3 dimensinos
396 with 900+ anchors
398 * Some clean up
400 2015-09-11 Andy Buckley <andy.buckley@cern.ch>
402 * Adding prof2-residuals script for ipol faithfulness testing.
404 * Rename scripts from prof-* to prof2-* to allow parallel installations of Prof1 and Prof2.
406 * Move some non-core scripts from bin to contrib.
408 * Make root-config dependency optional
410 * Lots of reworking to make prof-tune work again, and to use weight file parsing.
412 2015-09-10 Andy Buckley <andy.buckley@cern.ch>
414 * Use weight file path parsing in prof-ipol.
416 * Add prof.weights submodule, with a copy of PointMatcher and a
417 new read_pointmatchers function for handling weight files.
419 * Provide __version__attribute in the Python module.
421 * User script CLI consistency, simplification, general smoothing...
423 2015-08-30 Andy Buckley <andy.buckley@cern.ch>
425 * Add recursive scangrid generator function, and other tweaks.
427 2015-08-18 Andy Buckley <andy.buckley@cern.ch>
429 * Tweaks to ipol file parsing functions.
431 * Fix a bug in IpolBin's use of interpolated errors.
433 * Convert mean and median error parameterisation to use new 0th order ipols.
435 2015-08-15 Holger Schulz <holger.schulz@durham.ac.uk>
437 * A first version of catching singular matrices in the SVD
439 * Ipol has 1 additional argument, "threshold", that determines what
440 singular values are considered 0, fully propagated to pyext
442 * Going 2 Pro 2 Fessional
444 * Some startup checks of prof- scripts
446 * remove bin/prof-sampling, rename prof-prediction -> prof-predict
448 2015-08-13 Holger Schulz <holger.schulz@durham.ac.uk>
450 * Reverted unintended commit of hack.
452 * Added very basic prof-config script to bin
454 * Update PATH when sourcing setup.sh
456 2015-08-12 Holger Schulz <holger.schulz@durham.ac.uk>
458 * Added prof-prediction which reads in a ifile and a point in
459 parameter space (either a text file, a comma separated list or just
460 the args) and writes out an ipolhisto. Tested to work.
462 2015-08-03 Holger Schulz <holger.schulz@durham.ac.uk>
464 * Added 0-order polynomials, i.e. constant values.
465 Currently, the coefficent is simply the value of
466 the first anochorpoint.
468 2015-07-15 Holger Schulz <holger.schulz@durham.ac.uk>
470 * Some pull plotting functionality
472 2015-07-08 Holger Schulz <holger.schulz@durham.ac.uk>
474 * Calculate and print G.o.f.
476 * Write some meta info to results
478 * Read limits, fixed params from single textfile --limits
480 * Output files steered by --output
482 * Remove now obsolete bin/prof-interpolate
485 2015-07-08 Holger Schulz <holger.schulz@durham.ac.uk>
487 * First working version of prof-tune (pyminuit, simpleGOF)
489 * Multiprocessing to speed up prof-ipol (--multi)
491 * prof-ipol-tabulated to write out interpolation of tabulated data
493 * Minor bugfixes
496 2015-06-16 Andy Buckley <andy.buckley@cern.ch>
498 * Adding mean and median strategies for assigning constant ipol bin errors.
500 2015-06-03 Andy Buckley <andy.buckley@cern.ch>
502 * Adding a Bin base class in the Python side, and a coherent handling of multiple (interpolated) bin errors.
504 * Various tweaks and minor fixes.
506 2015-05-10 Andy Buckley <andy.buckley@cern.ch>
508 * Add 'Minuit' class importing to the __init__.py file.
510 2015-05-05 Andy Buckley <andy.buckley@cern.ch>
512 * Rewrite C++ ipolstring parsing using STL string functions rather than Boost: shorter! Boost dependency GONE :-)
514 * Replace boost::tuple with std::pair and clean up ParamPoints interface a bit.
516 * Add a --veto option to prof-sample, for a user-specified vetoing function.
518 * Move Sampler from prof-sample into professor2/__init__.py
520 2015-05-04 Andy Buckley <andy.buckley@cern.ch>
522 * Adding a first version of prof-tune with ipol reading from file.
524 * Adding options for (average) error interpolation and ipol persistency to prof-ipol and module functions.
526 2015-05-03 Andy Buckley <andy.buckley@cern.ch>
528 * Add bin/prof-ipol simple script using the functions below -- to be merged with prof-interpolate.
530 * professor2/__init__.py: Add Histo, DataBin, IpolBin and basic data handling functions.
532 * Remove unnecessary dlopen fiddling from professor2/__init__.py
534 2015-04-23 Andy Buckley <andy.buckley@cern.ch>
536 * Remove bound ParamPoints pointer from Ipol, and ditch lazy coeff evaluation in favour of simplicity and construct-time coeff calculating.
538 * Move long vector and coeff building into unbound functions rather than Ipol methods.
540 * Start of ChangeLog. Library has already iterated quite a bit.
542 * FILL IN EARLY HISTORY FROM HG LOG