Causality in Time Series - ClopiNet

More documents

Recommendations

Info

Guyon Statnikov Aliferis4. A Data RepositoryPart of our benchmarking effort is dedicated to collecting problems from diverse applicationdomains. Via the organization of competitions, we have successfully channeledthe effort or dozens of researchers to solve new problems of scientific and practicalinterest and identified effective methods. However, competition without collaborationis sterile. Recently, we have started introducing new dimensions to our effort of researchcoordination: stimulating creativity, collaborations, and data exchange. Weare organizing regular teleconference seminars. We have created a data repositoryfor the Causality Workbench already populated by 15 datasets. All the resources,which are the product of our effort, are freely available on the Internet at http://clopinet.com/causality. The repository already includes several time seriesdatasets, illustrating problems of practical and academic interest (see table 1):- Learning the structure of a fairly complex dynamic system that disobeys equilibrationmanipulationcommutability, and predicting the effect of manipulations that donot cause instabilities (the MIDS task) (Voortman et al., 2010);- Learning causal relationships using time series when noise is corrupting data in away that classical “Granger causality” fails (the NOISE task) (Nolte et al., 2010);- Uncovering which promotions affect most sales in a marketing database (thePROMO task) (Pellet, 2010);- Identifying in a manufacturing process (wafer production) faulty steps affectinga performance metric (the SEFTI task) (Tuv, 2010);- Modeling a biological signalling process (the SIGNET task) (Jenkins, 2010).The donor of the dataset NOISE (Guido Nolte) received the best dataset award. Thereviewers appreciated that the task includes both real data from EEG time series andartificial data modeling EEG. We want to encourage future data donors to move in thisdirection.5. Benchmarks and CompetitionsOur effort has been gaining momentum with the organization of two challenges, whicheach attracted over 50 participants. The first causality challenge we have organized(Causation and Prediction challenge, December 15 2007 - April 30 2008) allowed researchersboth from the causal discovery community and the machine learning communityto try their algorithms on sizable tasks of real practical interest in medicine, pharmacology,and sociology (see http://www.causality.inf.ethz.ch/challenge.php). The goal was to train models exclusively on observational data, then make predictionsof a target variable on data collected after intervention on the system understudy were performed. This first challenge reached a number of goals that we had setto ourselves: familiarizing many new researchers and practitioners with causal discoveryproblems and existing tools to address them, pointing out the limitations of current134
Causality WorkbenchTable 1: Time dependent datasets.“TP” is the data type, “NP” the number of participants who returned results and “V” the number ofviews as of January 2011. The semi-artificial datasets are obtained from simulators of real tasks. N is the number of variables,T is the number of time samples (not necessarily evenly spaced) and R the number of simulations with different initial statesor conditions.Name(TP; NP; V)MIDS(Artificial;NA; 794)NOISE(Real +artificial; NA;783)PROMO(Semiartificial;3;1601)SEFTI(Semiartificial;NA;908)SIGNET(Semi-artif.;2; 2663)Size Description ObjectiveT=12 sampled values in time(unevenly spaced); R=10000simulations. N=9 variables.Artificial: T=6000 time points;R=1000 simul.; N=2 var.Real: R=10 subjects.T≃200000 points sampled at256Hz. N=19 channels.T=365*3 days; R=1 simulation;N=1000 promotions + 100products.R=4000 manufacturing lots;T=300 async. operations (pairof values {one of N=25 toolIDs, date of proc.}) + cont.target (circuit perf. for each lot).T=21 asynchronous stateupdates; R=300 pseudodynamicsimulations; N=43 rules.Mixed Dynamic Systems. Simulatedtime-series based on linear Gaussianmodels with no latent common causes,but with multiple dynamic processes.Real and simulated EEG data.Learning causal relationships usingtime series when noise is corruptingdata causing the classical Grangercausality method to fail.Simulated marketing task. Dailyvalues of 1000 promotions and 100product sales for three yearsincorporating seasonal effects.Semiconductor manufacturing.Each wafer undergoes 300 steps eachinvolving one of 25 tools. A regressionproblem for quality control ofend-of-line circuit performance.Abscisic Acid Signaling Network.Model inspired by a true biologicalsignaling network.Use the training data tobuild a model able topredict the effects ofmanipulations on thesystem in test data.Artificial task: find thecausal dir. in pairs of var.Real task: Find whichbrain region influenceeach other.Predict a 1000x100boolean matrix of causalinfluences of promotionson product sales.Find the tools that areguilty of performancedegradation and eventualinteractions and influenceof time.Determine the set of 43boolean rules thatdescribe the network.135
Page 1:
Causality in Time SeriesVolume 5:Ca
Page 5:
PrefaceThe following is a print ver
Page 9 and 10:
JMLR: Workshop and Conference Proce
Page 11 and 12:
Linking Granger Causality and the P
Page 13 and 14:
Page 15 and 16:
Page 17 and 18:
Page 19 and 20:
Page 21 and 22:
Page 23 and 24:
Page 25 and 26:
Page 27 and 28:
Page 29 and 30:
Page 31 and 32:
Page 33 and 34:
Page 35 and 36:
Page 37 and 38:
Page 39:
Page 42 and 43:
Popescu1. IntroductionCausality is
Page 44 and 45:
Popescugeneral terms, if we are pre
Page 46 and 47:
PopescuType III error prob. γ = P
Page 48 and 49:
Popescuwhich would correspond to a
Page 50 and 51:
Popescuvalued vector. A Data Genera
Page 52 and 53:
Popescuy i =K∑︁A k y i−k + Bu
Page 54 and 55:
Popescuy i =K∑︁A k y i−k + Bu
Page 56 and 57:
PopescuFigure 1: SVAR causality and
Page 58 and 59:
Popescu6. Spectral methods and phas
Page 60 and 61:
PopescuH GCj→i|u = logD i − log
Page 62 and 63:
Popescu8.1. The cardinal transform
Page 64 and 65:
Popescuy N,i = Bx N,i (37)y = (1
Page 66 and 67:
Popescu(a) Unmixed colored noise(b)
Page 68 and 69:
PopescuAs we can see in both Figure
Page 70 and 71:
Popescuβ > 0 y C,i =x N,i =⎡K∑
Page 72 and 73:
Popescupretation of information flo
Page 74 and 75:
Popescuit is suggested that a princ
Page 76 and 77:
PopescuA. N. Kolmogorov and A. N. S
Page 78 and 79:
PopescuH. White and X. Lu. Granger
Page 80 and 81:
PopescuTable 6: α vs. ΨN * → 50
Page 82 and 83:
Roebroeck Seth Valdes-Sosaincreasin
Page 84 and 85:
Roebroeck Seth Valdes-SosaFigure 1:
Page 86 and 87:
Roebroeck Seth Valdes-Sosasignal fr
Page 88 and 89:
Roebroeck Seth Valdes-Sosablood flo
Page 90 and 91:
Roebroeck Seth Valdes-Sosaat lower
Page 92 and 93: Roebroeck Seth Valdes-SosaTable 1:
Page 94 and 95: Roebroeck Seth Valdes-Sosa1984) and
Page 96 and 97: Roebroeck Seth Valdes-Sosaanalysis
Page 98 and 99: Roebroeck Seth Valdes-Sosamarginal
Page 100 and 101: Roebroeck Seth Valdes-SosaThe initi
Page 102 and 103: Roebroeck Seth Valdes-Sosaobserved
Page 104 and 105: Roebroeck Seth Valdes-Sosaand corti
Page 106 and 107: Roebroeck Seth Valdes-SosaDaniel Co
Page 108 and 109: Roebroeck Seth Valdes-SosaM. Havlic
Page 110 and 111: Roebroeck Seth Valdes-SosaA. Roebro
Page 112 and 113: Roebroeck Seth Valdes-SosaV. Walsh
Page 114 and 115: Moneta Chlass Entner HoyerA traditi
Page 116 and 117: Moneta Chlass Entner Hoyerto the
Page 118 and 119: Moneta Chlass Entner Hoyersecond st
Page 120 and 121: Moneta Chlass Entner HoyerThe Wald
Page 122 and 123: Moneta Chlass Entner HoyerStep 3 Ap
Page 124 and 125: Moneta Chlass Entner Hoyerthe struc
Page 126 and 127: Moneta Chlass Entner Hoyerstricted
Page 128 and 129: Moneta Chlass Entner Hoyerwhere h 1
Page 130 and 131: Moneta Chlass Entner Hoyerlatter re
Page 132 and 133: Moneta Chlass Entner Hoyercausal mo
Page 134 and 135: Moneta Chlass Entner HoyerL. Baring
Page 136 and 137: Moneta Chlass Entner HoyerS. Johans
Page 138 and 139: Moneta Chlass Entner HoyerA. Yatche
Page 140 and 141: Guyon Statnikov Aliferisis extremel
Page 144 and 145: Guyon Statnikov Aliferismethods on
Page 146 and 147: Guyon Statnikov Aliferisvolume 3, p
show all

Causality in Time Series - ClopiNet

Create successful ePaper yourself

Delete template?

Save as template?