professor

view ChangeLog @ 538:28291b6236d1

Unbelievable, forgot to incude this file, sorry.
author Holger Schulz <holger.schulz@durham.ac.uk>
date Mon, 13 Nov 2017 08:32:25 -0600
parents a72d5522560d
children
line source
1 2017-11-10 Holger Schulz <iamholger@gmail.com>
3 * Release 2.2.2 --- this is the result of many months of fiddling
4 with especially multinest requirements.
6 * Sampling --- sobol, latin hypercube and random uniform sampler
7 --- something like PARAM 0 1 2 3 result in sub space
8 sampling
9 --- biases now work e.g. PARAM 0 1 exp(x)
10 * Ipol --- --order auto and friends automatically determine
11 best order, this avoids oversampling.
12 --- --logy parmeterises the logy of things, useful
13 for large paramspace and especially likelihood
14 --- --medianfilt can drop anchor points based on error
15 this is experimental and meant to deal with the
16 occasional huge NLO weight
17 --- The output has the valid region for each object now
18 stored at the end of the coefficients, this is also where
19 they are read from --- this allows the usage of
20 --medianfilt etc
21 * Tune --- Replace the grid sampling with the already existing
22 sampler machinery, so can use random, latin, sobol when
23 using --scan to find good start point for minuit
24 --- --profiles and --contours stores the minuit and minos
25 plots
26 --- limits are by default now set to the ipol region, this
27 can be overridden when using -x CL arg
28 --- tune script can read more than one ipol file
29 A lot of stuff has been put into libraries to allow reusing
30 individual steps.
32 * Multinest --- this is now supported in prof2-tune-nest
33 * GP --- gp_minimize via prof2-tune-gp
34 * emcee --- MCMC sampling via prof2-emcee
36 * Eigentunes --- prof2-eigentunes is back and working, deltaChi2
37 must be provided, however, via --target
39 * Bootstrapping in principle working but smearing is probably
40 more complicated than what we have now prof2-bootstrap
44 2017-08-22 Holger Schulz <holger.schulz@durham.ac.uk>
46 * Improvements to Ipol --- we allow different orders for each bin now,
47 there is also --order auto which does some jackknifing to figure out
48 the best suited polynomial order to avoid overfitting. The Ipol
49 outputformat is now binned 3 which dumps the minparam values and
50 maxparamvalues at the end of the coefficient string. This is necessary
51 as we now also allow to filter the inputs by median error (necessary
52 to tame NLO spiky histograms). The ranges are read back for each bin
53 separately now, happens in the C++ part.
55 * Overhaul of sampling, including things such as full support of sobol
56 sampling. Also, param range files with more than two numbers on a line
57 are interpreted as sub spaces for patch sampling.
59 * Add script prof2-ipolenvelopes which for each bin finds the minimum
60 and maximum polynomial value (using migrad) in the range of the
61 ipols validity. This is nice especially to check envelopes once
62 filtering in the ipol step was done
64 2017-05-18 Holger Schulz <holger.schulz@durham.ac.uk>
66 * Refitting of ipols using smeared ipol output.
67 Required some modification in data io structure.
69 2017-05-04 Holger Schulz <holger.schulz@durham.ac.uk>
71 * Fix in prof2-tune when checking analysis names, at the same time
72 allow for for closure test tuning, i.e. tune against input folder.
75 2017-05-02 Holger Schulz <holger.schulz@durham.ac.uk>
77 * Use -march=native according to eigen3 faq for better performance
78 as well as -O3 in pyext
80 * Replace modulus based status message by simple counter in ipol stage
82 * Set version to 2.2.2beta for imminent release
84 2017-04-28 Holger Schulz <holger.schulz@durham.ac.uk>
86 * Use BDSSVD in case Eigen >=3.3.3
88 * Reworked prof2-ipol and ipol.py to use mk_ipolbin rather than
89 mk_ipolhisto, some speedups plus additional benefit of being
90 able to fiddle with inividual bins in the ipol building stage
92 * add --minos switch to prof2-tune toi run MINOS
94 * Write GOF and NDOF into result file
96 2017-02-21 Andy Buckley <andy.buckley@cern.ch>
98 * Change prof2-ipol ierr mode default to 'symm'
100 2017-02-17 Andy Buckley <andy.buckley@cern.ch>
102 * Add a --limit-errs argument for prof2-tune, to avoid annoying
103 re-reading of large run dirs
105 * Tweaks to default error ipol order, and release figure handles
106 in envelope plotting.
108 2017-02-02 Holger Schulz <holger.schulz@durham.ac.uk>
110 * Update Copyright string
112 * Instead of exiting when prof2-ipol encounters nan coefficients in
113 value or error ipols, spit out a warning and ignore that histogram
114 entirely --- this makes prof2-ipol and friends much more robust
116 * tag as 2.2.1 and release.
118 2017-02-01 Holger Schulz <holger.schulz@durham.ac.uk>
120 * Use Agg as default matplotlib backend.
122 * Bump version to 2.2.1
124 2016-12-19 Holger Schulz <holger.schulz@cern.ch>
126 * Even more robust prof-tune, added --filter CL option to
127 automatically get rid of zero error bins
129 * Fixed a bug where IpolBin.n was never set which is needed for
130 the PointMatcher (i.e. weights syntax)
132 * Not using YODA Pointmatcher but own due to different regexes
133 required.
135 * Make release 2.2
137 2016-12-15 Holger Schulz <holger.schulz@cern.ch>
139 * More robust scripts and more meaningful error messages, especially
140 prof2-tune
142 2016-11-30 Holger Schulz <holger.schulz@cern.ch>
144 * Replace Counter code when running mk_Structure with much faster code
146 * Bump version to 2.2
148 * Bump DataFormat to binned 2
150 * Allow passing file names or open files (or similar streams) to
151 ipolio functions
153 * prof2-ipol can read yaml now
155 * Add some of Elliot's code for choosing sub spaces in prof2-ipol and
156 some plot routines to contrib
158 * Add SOBOL sampling to contrib
160 2016-08-26 Holger Schulz <holger.schulz@cern.ch>
162 * Use more readable Cython DL path in bootstrap
164 2016-07-18 Andy Buckley <andy.buckley@cern.ch>
166 * Typo fixes -- thanks to Leif Gellersen.
168 2016-07-14 Andy Buckley <andy.buckley@cern.ch>
170 * Release version 2.1.4
172 * Improve Makefile tests for ROOT and Cython versions.
174 2016-07-11 Andy Buckley <andy.buckley@cern.ch>
176 * Add Ipol min/max param vals getters and map them into Python.
178 * Add a built-in /REF-stripping option to data histo loading.
180 * Improvements to the Python ipol metadata interface.
182 2016-07-05 Andy Buckley <andy.buckley@cern.ch>
184 * Try to optimise long-vector building by not executing pow() if the exponent would be zero.
186 * Split C++ calcValue function into two, to allow re-use of a precalculated long-vector.
188 2016-06-23 Andy Buckley <andy.buckley@cern.ch>
190 * Split 'dot product' part of ipol value calculation into a standalone calcValue function.
192 * Use const references when passing const vectors in C++ API.
194 2016-04-14 Holger Schulz <holger.schulz@cern.ch>
196 * Introduce 1D sensitivity based on gradient
198 * User needs to specify --cmap or --grad when running prof2-sens now
200 2016-04-01 Holger Schulz <holger.schulz@cern.ch>
202 * Add a first version of the parametrisation using Machine Learning
204 2016-02-16 Holger Schulz <holger.schulz@cern.ch>
206 * Bump to version 2.1.3
208 2016-02-15 Holger Schulz <holger.schulz@cern.ch>
210 * prof2-sens: bugfix, add weight file usage, clean up CL
212 * prof2-I: add weight file usage
214 * prof2-predict: add weight file usage, fix output writing
216 * Extended documentation
218 2016-02-12 Holger Schulz <holger.schulz@cern.ch>
220 * Documentation including asciinema
222 2016-02-11 Holger Schulz <holger.schulz@cern.ch>
224 * Remove min_runs function, replace with numCoeffs
225 from C++ code
227 * Add prof2-envelopes script
229 * In prof-tune simplechi2 --- continue on weight==0
230 to allow some robustness when switching off bins
233 2016-02-10 Holger Schulz <holger.schulz@cern.ch>
235 * First version of prof2-runcombs plus sampling functions
236 salvaged from Prof1, modified to have no scipy dependence
238 * Add --rc switch to prof2-ipol to optionally read runcombs
239 Syntax is --rc runcoms.dat[colon]5
240 (Minimally invasive addon)
242 * Code cull in prof2-ipol. Only support multiprocessing
243 version, default 1 thread. Changed CL switch to '-j'.
246 2015-12-16 Holger Schulz <holger.schulz@cern.ch>
248 * Only calculate polynomials structure once per instance
249 -> significant speed increase
251 * Add the brute force grid scan (--scan) to prof2-tune
254 2015-12-11 Holger Schulz <holger.schulz@cern.ch>
256 * Minimiser tweaks --- no more need for param translation,
257 make strategy steerable (--strategy or -s), -q suppresses
258 iminuit output now, calculation and writing of correlation
259 matrix into results, dump weights into results file
261 2015-12-10 Holger Schulz <holger.schulz@cern.ch>
263 * Undo ui madness
265 * Remove numpy dependence (except sampling.py)
267 * Add prof-sens for sensitivity plotting exploting derivatives
269 2015-12-09 Holger Schulz <holger.schulz@cern.ch>
271 * Can specify PREFIX when calling make
273 2015-12-09 Andy Buckley <andy.buckley@cern.ch>
275 * Improve prof2-predict UI.
277 * Add metadata printing in prof2-lsipol debug mode.
279 * Add -v and -q flag shortcuts to all scripts.
281 2015-12-07 Holger Schulz <holger.schulz@cern.ch>
283 * Use __all__ in __init__.py
285 * Started sphinx documentation
287 * Support only iminuit, bump version to 2.2.alpha, restructure imports
289 * Sanity checks when calling parametrisation
291 2015-12-03 Andy Buckley <andy.buckley@cern.ch>
293 * Provide a version() function in the C++ library, and map it into
294 Python. Both sources use the VERSION variable defined in the
295 Makefile at build time.
297 * Reduce SVD fitting threshold to 1e-20.
299 2015-12-03 Holger Schulz <holger.schulz@cern.ch>
301 * Make derivative calculation work also when scaling params
303 * Let YODA write YODA
305 2015-12-02 Holger Schulz <holger.schulz@cern.ch>
307 * Derivative calculation
309 2015-11-28 Andy Buckley <andy.buckley@cern.ch>
311 * Version 2.1.2
313 * Revert range shift in Ipol.cc, since it produced (small)
314 numerical errors rather than improved stability. To be
315 understood...
317 * Add test/mkpolydata script, for generating polynomial pseudodata
318 used in closure tests.
320 2015-11-27 Andy Buckley <andy.buckley@cern.ch>
322 * Fix bug in histo loading.
324 * Adding log binning and other options to prof2-residuals.
326 2015-11-24 Andy Buckley <andy.buckley@cern.ch>
328 * Version 2.1.1 release.
330 * Fix prof2-predict YODA writing, and improve DataHisto.toYODA.
332 * Change parameter rescaling from using the [0,1] range to instead
333 use [1,2], to avoid arbitrarily scaled param numbers. BREAKS IPOL
334 FILE COMPATIBILITY WITH 2.1.0!!
336 * Set minimizer initial step size = prange/10.
338 * Add automatic determination of maximum MC errors in prof-tune,
339 and use them to regularize interpolated errs which could distort
340 the fit.
342 * Separate internal Minuit function names (with syntax
343 restrictions) from the free-form external param names.
345 2015-11-23 Andy Buckley <andy.buckley@cern.ch>
347 * Add pyext/professor2/chi2.py
349 * Add find_maxerrs() to the Python interface, for use in error regularisation.
351 * Use the param file name as a regex.
353 * Use __slots__ in the internal histogram and bin types, for better memory efficiency.
355 * Fix (re-fix?) use of inappropriate indices in calculation of mean and median MC uncertainties.
357 * More proactive data object deletion after conversion to YODA data format.
359 2015-11-22 Andy Buckley <andy.buckley@cern.ch>
361 * Version 2.1.0 (significant version change due to scaling, all-orders, and ROOT support)
363 2015-11-18 Andy Buckley <andy.buckley@cern.ch>
365 * dataio.py: Add ROOT file reading ability.
367 * Change rescaling I/O behaviour a bit, moving some hacked
368 Python-only Ipol string constructors into explicit API improvements, and
369 only activating rescaling if DoParamScaling is set true.
371 2015-11-16 Andy Buckley <andy.buckley@cern.ch>
373 * Add CPPFLAGS steering in Makefile and API tidying.
375 2015-11-11 Holger Schulz <holger.schulz@cern.ch>
377 * Add automatic raw param value -> unit range mapping to Ipol.
379 * Add any-order long vector calculation.
381 2015-11-10 Andy Buckley <andy.buckley@cern.ch>
383 * Add prof2-lsipol as a handy way to make a tuning weights file
384 and for general convenience.
386 2015-10-06 Andy Buckley <andy.buckley@cern.ch>
388 * Version 2.0.0
390 * Add 'make dist' target to Makefile, and tweak/reduce Cython necessity
392 * Allow calls to IpolBin and IpolHisto methods with params passed
393 as unpacked *args as well as a packed iterable.
395 * Add possibility to pass params as an dict-like as well as a list
396 or tuple of floats -- note that the dict-like must return
397 *ordered* values consistent with the Ipol training. Also ensure
398 the float type.
400 * Add conversion of C++ exceptions to Python exceptions.
402 * Add professor2.utils module, providing an opportunistic use of
403 OrderedDict for params so the params order required by Ipol is the
404 same as the order in which they are specified in the (first run's)
405 params.dat file.
407 * Change Python-mapped Ipol methods with no args to be properties.
409 * Add nice __repr__ methods for Histo and Bin objects.
411 * Specialise Histo as IpolHisto and DataHisto, and provide
412 toData{Histo,Bin} on IpolHisto and IpolBin respectively.
414 * Remove ProfMaster.
416 2015-10-04 Andy Buckley <andy.buckley@cern.ch>
418 * Add protection of svd.setThreshold call to ensure it is
419 supported/needed, via Eigen version number macros.
421 * Add vmin and vmax args to Ipol and IpolBin in Python.
423 2015-10-02 Andy Buckley <andy.buckley@cern.ch>
425 * Put histo file reading in a try..except block so read failures
426 on non-histo files in the run dirs are not fatal.
428 * Add protection in the histo loader, so we don't trip up on
429 Rivet's new cross-section and counter objects.
431 2015-09-30 Andy Buckley <andy.buckley@cern.ch>
433 * src/Ipol.h: Use Eigen/SVD include path rather than non-standard eigen3/Eigen/SVD.
435 2015-09-14 Holger Schulz <holger.schulz@durham.ac.uk>
437 * Use Eigen's svd.setThreshold(1e-20) to prevent the errors we saw with Eigen
438 3.2.2 and later when looking at 5th order polynomials in 3 dimensinos
439 with 900+ anchors
441 * Some clean up
443 2015-09-11 Andy Buckley <andy.buckley@cern.ch>
445 * Adding prof2-residuals script for ipol faithfulness testing.
447 * Rename scripts from prof-* to prof2-* to allow parallel installations of Prof1 and Prof2.
449 * Move some non-core scripts from bin to contrib.
451 * Make root-config dependency optional
453 * Lots of reworking to make prof-tune work again, and to use weight file parsing.
455 2015-09-10 Andy Buckley <andy.buckley@cern.ch>
457 * Use weight file path parsing in prof-ipol.
459 * Add prof.weights submodule, with a copy of PointMatcher and a
460 new read_pointmatchers function for handling weight files.
462 * Provide __version__attribute in the Python module.
464 * User script CLI consistency, simplification, general smoothing...
466 2015-08-30 Andy Buckley <andy.buckley@cern.ch>
468 * Add recursive scangrid generator function, and other tweaks.
470 2015-08-18 Andy Buckley <andy.buckley@cern.ch>
472 * Tweaks to ipol file parsing functions.
474 * Fix a bug in IpolBin's use of interpolated errors.
476 * Convert mean and median error parameterisation to use new 0th order ipols.
478 2015-08-15 Holger Schulz <holger.schulz@durham.ac.uk>
480 * A first version of catching singular matrices in the SVD
482 * Ipol has 1 additional argument, "threshold", that determines what
483 singular values are considered 0, fully propagated to pyext
485 * Going 2 Pro 2 Fessional
487 * Some startup checks of prof- scripts
489 * remove bin/prof-sampling, rename prof-prediction -> prof-predict
491 2015-08-13 Holger Schulz <holger.schulz@durham.ac.uk>
493 * Reverted unintended commit of hack.
495 * Added very basic prof-config script to bin
497 * Update PATH when sourcing setup.sh
499 2015-08-12 Holger Schulz <holger.schulz@durham.ac.uk>
501 * Added prof-prediction which reads in a ifile and a point in
502 parameter space (either a text file, a comma separated list or just
503 the args) and writes out an ipolhisto. Tested to work.
505 2015-08-03 Holger Schulz <holger.schulz@durham.ac.uk>
507 * Added 0-order polynomials, i.e. constant values.
508 Currently, the coefficent is simply the value of
509 the first anochorpoint.
511 2015-07-15 Holger Schulz <holger.schulz@durham.ac.uk>
513 * Some pull plotting functionality
515 2015-07-08 Holger Schulz <holger.schulz@durham.ac.uk>
517 * Calculate and print G.o.f.
519 * Write some meta info to results
521 * Read limits, fixed params from single textfile --limits
523 * Output files steered by --output
525 * Remove now obsolete bin/prof-interpolate
528 2015-07-08 Holger Schulz <holger.schulz@durham.ac.uk>
530 * First working version of prof-tune (pyminuit, simpleGOF)
532 * Multiprocessing to speed up prof-ipol (--multi)
534 * prof-ipol-tabulated to write out interpolation of tabulated data
536 * Minor bugfixes
539 2015-06-16 Andy Buckley <andy.buckley@cern.ch>
541 * Adding mean and median strategies for assigning constant ipol bin errors.
543 2015-06-03 Andy Buckley <andy.buckley@cern.ch>
545 * Adding a Bin base class in the Python side, and a coherent handling of multiple (interpolated) bin errors.
547 * Various tweaks and minor fixes.
549 2015-05-10 Andy Buckley <andy.buckley@cern.ch>
551 * Add 'Minuit' class importing to the __init__.py file.
553 2015-05-05 Andy Buckley <andy.buckley@cern.ch>
555 * Rewrite C++ ipolstring parsing using STL string functions rather than Boost: shorter! Boost dependency GONE :-)
557 * Replace boost::tuple with std::pair and clean up ParamPoints interface a bit.
559 * Add a --veto option to prof-sample, for a user-specified vetoing function.
561 * Move Sampler from prof-sample into professor2/__init__.py
563 2015-05-04 Andy Buckley <andy.buckley@cern.ch>
565 * Adding a first version of prof-tune with ipol reading from file.
567 * Adding options for (average) error interpolation and ipol persistency to prof-ipol and module functions.
569 2015-05-03 Andy Buckley <andy.buckley@cern.ch>
571 * Add bin/prof-ipol simple script using the functions below -- to be merged with prof-interpolate.
573 * professor2/__init__.py: Add Histo, DataBin, IpolBin and basic data handling functions.
575 * Remove unnecessary dlopen fiddling from professor2/__init__.py
577 2015-04-23 Andy Buckley <andy.buckley@cern.ch>
579 * Remove bound ParamPoints pointer from Ipol, and ditch lazy coeff evaluation in favour of simplicity and construct-time coeff calculating.
581 * Move long vector and coeff building into unbound functions rather than Ipol methods.
583 * Start of ChangeLog. Library has already iterated quite a bit.
585 * FILL IN EARLY HISTORY FROM HG LOG