Fork me on GitHub

source: git/external/fastjet/contribs/RecursiveTools/README@ aa3c287

ImprovedOutputFile Timing dual_readout llp 3.4.2pre16
Last change on this file since aa3c287 was b7b836a, checked in by Pavel Demin <pavel-demin@…>, 6 years ago

update FastJet library to 3.3.1 and FastJet Contrib library to 1.036

  • Property mode set to 100644
File size: 13.6 KB
RevLine 
[1f1f858]1------------------------------------------------------------------------
2RecursiveTools FastJet contrib
3------------------------------------------------------------------------
4
5The RecursiveTools FastJet contrib aims to provide a common contrib
6for a number of tools that involve recursive reclustering/declustering
7of a jet for tagging or grooming purposes.
8
9Currently it contains:
10
11- ModifiedMassDropTagger
12 This corresponds to arXiv:1307.0007 by Mrinal Dasgupta, Alessandro
13 Fregoso, Simone Marzani and Gavin P. Salam
14
15- SoftDrop
16 This corresponds to arXiv:1402.2657 by Andrew J. Larkoski, Simone
17 Marzani, Gregory Soyez, Jesse Thaler
18
[b7b836a]19- RecursiveSoftDrop
20- BottomUpSoftDrop
21 This corresponds to arXiv:1804.03657 by Frederic Dreyer, Lina
22 Necib, Gregory Soyez and Jesse Thaler
23
24- IteratedSoftDrop
25 This corresponds to arXiv:1704.06266 by Christopher Frye, Andrew J.
26 Larkoski, Jesse Thaler, Kevin Zhou
27
[1f1f858]28- Recluster
29 A generic tool to recluster a given jet into subjets
[b7b836a]30 Note: a Recluster class is available natively in FastJet since v3.1.
31 Users are therefore encouraged to use the FastJet version
32 rather than this one which is mostly provided for
33 compatibility of this contrib with older versions of FastJet.
[1f1f858]34
35The interface for these tools is described in more detail below, with
36all of the available options documented in the header files.
37
38One note about nomenclature. A groomer is a procedure that takes a
39PseudoJet and always returns another (non-zero) PseudoJet. A tagger is
40a procedure that takes a PseudoJet, and either returns another PseudoJet
41(i.e. tags it) or returns an empty PseudoJet (i.e. doesn't tag it).
42
43------------------------------------------------------------------------
44ModifiedMassDropTagger
45------------------------------------------------------------------------
46
47The Modified Mass Drop Tagger (mMDT) recursively declusters a jet,
48following the largest pT subjet until a pair of subjets is found that
49satisfy the symmetry condition on the energy sharing
50
51 z > z_cut
52
53where z_cut is a predetermined value. By default, z is calculated as
54the scalar pT fraction of the softest subjet. Note that larger values
55of z_cut correspond to a more restrictive tagging criteria.
56
57By default, mMDT will first recluster the jet using the CA clustering
58algorithm, which means that mMDT can be called on any jet, regardless
59of the original jet finding measure.
60
61A default mMDT can be created via
62
63 double z_cut = 0.10;
64 ModifiedMassDropTagger mMDT(z_cut);
65
66More options are available in the full constructor. To apply mMDT,
67one simply calls it on the jet of interest.
68
69 PseudoJet tagged_jet = mMDT(original_jet);
70
71Note that mMDT is a tagger, such that tagged_jet will only be non-zero
72if the symmetry cut z > z_cut is satisfied by some branching of the
73clustering tree.
74
75To gain additional information about the mMDT procedure, one can use
76
77 tagged_jet.structure_of<ModifiedMassDropTagger>()
78
79which gives access to information about the delta_R between the tagged
80subjets, their z value, etc.
81
82------------------------------------------------------------------------
83SoftDrop
84------------------------------------------------------------------------
85
86The SoftDrop procedure is very similar to mMDT, albeit with a
[b7b836a]87generalised symmetry condition:
[1f1f858]88
89 z > z_cut * (R / R0)^beta
90
91Note that larger z_cut and smaller beta correspond to more aggressive
92grooming of the jet.
93
94SoftDrop is intended to be used as a groomer (instead of as a tagger),
95such that if the symmetry condition fails throughout the whole
96clustering tree, SoftDrop will still return a single particle in the
97end. Apart from the tagger/groomer distinction, SoftDrop with beta=0 is
98the same as mMDT.
99
100A default SoftDrop groomer can be created via:
101
102 double z_cut = 0.10;
103 double beta = 2.0;
104 double R0 = 1.0; // this is the default value
105 SoftDrop sd(z_cut,beta,R0);
106
107and acts on a desired jet as
108
109 PseudoJet groomed_jet = sd(original_jet);
110
111and additional information can be obtained via
112
113 groomed_jet.structure_of<SoftDrop>()
114
115SoftDrop is typically called with beta > 0, though beta < 0 is still a
116viable option. Because beta < 0 is infrared-collinear unsafe in
117grooming mode, one probably wants to switch to tagging mode for negative
118beta, via set_tagging_mode().
119
[b7b836a]120------------------------------------------------------------------------
121RecursiveSoftDrop
122------------------------------------------------------------------------
123
124The RecursiveSoftDrop procedure applies the Soft Drop procedure N times
125in a jet in order to find up to N+1 prongs. N=0 makes no modification
126to the jet, and N=1 is equivalent to the original SoftDrop.
127
128Once one has more than one prong, one has to decide which will be
129declustered next. At each step of the declustering procedure, one
130undoes the clustering which has the largest declustering angle
131(amongst all the branches that are searched for substructure). [see
132"set_fixed_depth" below for an alternative]
133
134Compared to SoftDrop, RecursiveSoftDrop takes an extra argument N
135specifying the number of times the SoftDrop procedure is recursively
136applied. Negative N means that the procedure is applied until no
137further substructure is found (i.e. corresponds to taking N=infinity).
138
139 double z_cut = 0.10;
140 double beta = 2.0;
141 double R0 = 1.0; // this is the default value
142 int N = -1;
143 RecursiveSoftDrop rsd(z_cut, beta, N, R0);
144
145One then acts on a jet as
146
147 PseudoJet groomed_jet = rsd(jet)
148
149and get additional information via
150
151 groomed_jet.structure_of<RecursiveSoftDrop>()
152
153------------------------------------------------------------------------
154IteratedSoftDrop
155------------------------------------------------------------------------
156
157Iterated Soft Drop (ISD) is a repeated variant of SoftDrop. After
158performing the Soft Drop procedure once, it logs the groomed symmetry
159factor, then recursively performs Soft Drop again on the harder
160branch. This procedure is repeated down to an (optional) angular cut
161theta_cut, yielding a set of symmetry factors from which observables
162can be built.
163
164An IteratedSoftDrop tool can be created as follows:
165
166 double beta = -1.0;
167 double z_cut = 0.005;
168 double theta_cut = 0.0;
169 double R0 = 0.5; // characteristic radius of jet algorithm
170 IteratedSoftDrop isd(beta, z_cut, double theta_cut, R0);
171
172By default, ISD applied on a jet gives a result of type
173IteratedSoftDropInfo that can then be probed to obtain physical
174observables
175
176 IteratedSoftDropInfo isd_info = isd(jet);
177
178 unsigned int multiplicity = isd_info.multiplicity();
179 double kappa = 1.0; // changes angular scale of ISD angularity
180 double isd_width = isd_info.angularity(kappa);
181 vector<pair<double,double> > zg_thetags = isd_info.all_zg_thetag();
182 vector<pair<double,double> > zg_thetags = isd_info();
183 for (unsigned int i=0; i< isd_info.size(); ++i){
184 cout << "(zg, theta_g)_" << i << " = "
185 << isd_info[i].first << " " << isd_info[i].second << endl;
186 }
187
188Alternatively, one can directly get the multiplicity, angularity, and
189(zg,thetag) pairs from the IteratedSoftDrop class, at the expense of
190re-running the declustering procedure:
191
192 unsigned int multiplicity = isd.multiplicity(jet);
193 double isd_width = isd.angularity(jet, 1.0);
194 vector<pair<double,double> > zg_thetags = isd.all_zg_thetag(jet);
195
196
197Note: the iterative declustering procedure is the same as what one
198 would obtain with RecursiveSoftDrop with an (optional) angular cut
199 and recursing only in the hardest branch [see the "Changing
200 behaviour" section below for details], except that it returns some
201 information about the jet instead of a modified jet as RSD does.
202
203
204------------------------------------------------------------------------
205BottomUpSoftDrop
206------------------------------------------------------------------------
207
208This is a bottom-up version of the RecursiveSoftDrop procedure, in a
209similar way as Pruning can be seen as a bottom-up version of Trimming.
210
211In practice, the jet is reclustered and at each step of the clustering
212one checks the SoftDrop condition
213
214 z > z_cut * (R / R0)^beta
215
216If the condition is met, the pair is recombined. If the condition is
217not met, only the hardest of the two objects is kept for further
218clustering and the softest is rejected.
219
[1f1f858]220------------------------------------------------------------------------
221Recluster
222------------------------------------------------------------------------
223
[b7b836a]224 *** NOTE: this is provided only for backwards compatibility ***
225 *** with FastJet <3.1. For FastJet >=3.1, the native ***
226 *** fastjet::Recluster is used instead ***
227
[1f1f858]228The Recluster class allows the constituents of a jet to be reclustered
229with a different recursive clustering algorithm. This is used
[b7b836a]230internally in the mMDT/SoftDrop/RecursiveSoftDrop/IteratedSoftDrop
231code in order to recluster the jet using the CA algorithm. This is
232achieved via
[1f1f858]233
234 Recluster ca_reclusterer(cambridge_algorithm,
235 JetDefinition::max_allowable_R);
236 PseudoJet reclustered_jet = ca_reclusterer(original_jet);
237
238Note that reclustered_jet creates a new ClusterSequence that knows to
239delete_self_when_unused.
240
241------------------------------------------------------------------------
242Changing behaviour
243------------------------------------------------------------------------
244
[b7b836a]245The behaviour of the all the tools provided here
246(ModifiedMassDropTagger, SoftDrop, RecursiveSoftDrop and
247IteratedSoftDrop) can be tweaked using the following options:
[1f1f858]248
[b7b836a]249SymmetryMeasure = {scalar_z, vector_z, y, theta_E, cos_theta_E}
250 [constructor argument]
[1f1f858]251 : The definition of the energy sharing between subjets, with 0
[b7b836a]252 corresponding to the most asymmetric.
253 . scalar_z = min(pt1,pt2)/(pt1+pt2) [default]
254 . vector_z = min(pt1,pt2)/pt_{1+2}
255 . y = min(pt1^2,pt2^2)/m_{12}^2 (original y from MDT)
256 . theta_E = min(E1,E2)/(E1+E2),
257 with angular measure theta_{12}^2
258 . cos_theta_E = min(E1,E2)/(E1+E2),
259 with angular measure 2[1-cos(theta_{12})]
260 The last two variants are meant for use in e+e- collisions,
261 together with the "larger_E" recursion choice (see below)
262
263RecursionChoice = {larger_pt, larger_mt, larger_m, larger_E}
264 [constructor argument]
[1f1f858]265 : The path to recurse through the tree after the symmetry condition
[b7b836a]266 fails. Options refer to transverse momentum (pt), transverse mass
267 (mt=sqrt(pt^2+m^2), mass (m) or energy (E). the latter is meant
268 for use in e+e- collisions
[1f1f858]269
270mu_cut [constructor argument]
271 : An optional mass drop condition
272
[b7b836a]273set_subtractor(subtractor*) [or subtractor as a constructor argument]
[1f1f858]274 : provide a subtractor. When a subtractor is supplied, the
275 kinematic constraints are applied on subtracted 4-vectors. In
276 this case, the result of the ModifiedMassDropTagger/SoftDrop is a
277 subtracted PseudoJet, and it is assumed that
278 ModifiedMassDropTagger/SoftDrop is applied to an unsubtracted jet.
279 The latter default can be changed by calling
280 set_input_jet_is_subtracted().
281
282set_reclustering(bool, Recluster*)
283 : An optional setting to recluster a jet with a different jet
284 recursive jet algorithm. The code is only designed to give sensible
285 results with the CA algorithm, but other reclustering algorithm
286 (especially kT) may be appropriate in certain contexts.
287 Use at your own risk.
288
289set_grooming_mode()/set_tagging_mode()
290 : In grooming mode, the algorithm will return a single particle if the
291 symmetry condition fails for the whole tree. In tagging mode, the
292 algorithm will return an zero PseudoJet if no symmetry conditions
293 passes. Note that ModifiedMassDropTagger defaults to tagging mode
294 and SoftDrop defaults to grooming mode.
295
[b7b836a]296set_verbose_structure(bool)
297 : when set to true, additional information will be stored in the jet
298 structure. This includes in particular values of symmetry,
299 delta_R, and mu of dropped branches
300
301For the specific case of RecursiveSoftDrop, additional tweaking is
302possible via the following methods
303
304set_fixed_depth_mode(bool)
305 : when this is true, RSD will recurse (N times) into all the
306 branches found during the previous iteration [instead of recursing
307 through the largest declustering angle until N prongs have been
308 found]. This yields at most 2^N prong. For infinite N, the two
309 options are equivalent.
310
311set_dynamical_R0(bool)
312 : By default the angles in the SD condition are normalised to the
313 parameter R0. With "dynamical R0", RSD will dynamically adjust R0
314 to be the angle between the two prongs found during the previous
315 iteration.
316
317set_hardest_branch_only(bool)
318 : When substructure is found, only recurse into the hardest of the
319 two branches for further substructure search. This uses the class
320 RecursionChoice.
321
322set_min_deltaR_squared(double):
323 : set a minimal angle (squared) at which we stop the declustering
324 procedure. This cut is ineffective for negative values of the
325 argument.
326
[1f1f858]327------------------------------------------------------------------------
328Technical Details
329------------------------------------------------------------------------
330
331Both ModifiedMassDropTagger and SoftDrop inherit from
332RecursiveSymmetryCutBase, which provides a common codebase for recursive
333declustering of a jet with a symmetry cut condition. A generic
334RecursiveSymmetryCutBase depends on the following (virtual) functions
335(see header file for exact full specs, including constness):
336
337double symmetry_cut_fn(PseudoJet &, PseudoJet &)
338 : The function that defines the symmetry cut. This is what actually
339 defines different recursive declustering schemes, and all classes
340 that inherit from RecursiveSymmetryCutBase must define this
341 function.
342
343string symmetry_cut_description()
344 : the string description of the symmetry cut.
345
346------------------------------------------------------------------------
Note: See TracBrowser for help on using the repository browser.