SlideShare ist ein Scribd-Unternehmen logo
1 von 6
Downloaden Sie, um offline zu lesen
Supplementary Methods
To determine whether a time series reflects linear or nonlinear processes we
compare the out-of-sample forecast skill of a linear model versus an equivalent nonlinear
model. To do this, we apply a two-step procedure: 1) we use simplex-projection1
to
identify the best embedding dimension, and 2) we use this embedding in the S-map
procedure2
to assess the nonlinearity of the time series. In both cases, model performance
is evaluated out-of-sample with the time series divided into equal halves. The first half
(library set, X) is used to build the model, while the second half (prediction set, Y) is
reserved to judge the out-of-sample performance of model forecasts originating from the
library. This forecast protocol is a rigorous standard that avoids model over-fitting or
arbitrary fits to the data.
Simplex projection
Simplex projection is a nearest-neighbor forecasting algorithm that involves tracking
the forward evolution of nearby points in an embedding (a lagged coordinate state space
reconstruction).
3, 4, 5
Thus, similar past events are used to forecast the future, with the
important caveat that the dimensionality of the embedding determines what past events
are similar (nearby) to the predictee.
Given a library set of n points, we generate an E-dimensional embedding from the
time series data using lagged coordinates so that the library vectors
Xt , {xt , xt-, xt-2 … xt-(E-1)} and prediction vectors Yt , {yt , yt-, yt-2 … yt-(E-1)} are points
in E-dimensional state space (where =1, and t  [1, 2, … , n] ). We then choose the E+1
neighbors of Yt from the library set X that form the vertices of the smallest simplex that
contains Yt as an interior point. The forecast is based on how these nearby library points
(domain simplex) move forward in time (range simplex). Again, the key is in the
definition of “nearby” which depends critically on the dimension of the embedding. Thus,
the domain simplex is projected forward into its range by incrementing the time index of
each neighbor in the domain simplex by 1. The forecast is simply a weighted average of
the values given by the range simplex. The weights depend exponentially on proximity of
the library vectors in the domain simplex to the predictee vector Yt (illustrated in
Supplementary Figure 1). For this study an exploratory series of embedding dimensions
(E) ranging from 1 to 20 (or higher) are used to evaluate the prediction, and the best E is
chosen based on prediction skill. This embedding is then used in the S-map procedure.
S-map algorithm
S-maps are an extension of standard linear autoregressive models in which the
coefficients depend on the location of the predictee Yt in an E-dimensional embedding.
New coefficients are recalculated (from the library set X) by singular value
decomposition (SVD) for each new prediction. In this calculation, the weight given to
each vector in the library depends on how close that vector Xt is to the predictee Yt. The
extent of this weighting is determined by the parameter .
As above, we generate an E-dimensional embedding from points in the library using
lagged coordinates to obtain an embedded time series with vectors Xt  RE+1
, where Xt(0)
= 1 is the constant term in the solution of Eq. (S2) below. Let the time series
observation in the prediction set Tp time steps forward be Yt+Tp(1)=Y(t).
Then the forecast for Y(t) is

ˆYt  Ct ( j)Xt ( j)
j 0
E
 (S1)
For our analysis, we chose TP = 1. For each E-dimensional predictee vector Yt, C is
solved by SVD using the library set as follows:
B = AC, (S2)
where

Bi (|| Xi Yt ||)Yi,

Aij (|| Xi Yt ||)Xi( j), and

(d)  edit /d


 0, dit is the distance between Yt and the ith
neighbor vector Xi in the library
embedding, and the scale vector,

d , is the average distance between neighbors in the
library. Note that A has dimensions

n (E 1), where

n = size of the library. Again, a
different map is generated for each forecast, with the weightings in each map depending
on the location of the predictee in the E-dimensional state space. This weighting
procedure is governed by the tuning parameter , where  = 0 gives a global linear map,
and increasing values of  give increasingly local or nonlinear mappings. Note that when
 = 0, all vectors are weighted equally so a single (global) linear map can be used for all
predictions. In the case where  > 0, vectors closest to the predictee in state-space are
weighted more heavily in the SVD solution. Such forecasts emphasize local information
in the library set, and are therefore nonlinear. The method is illustrated in the
Supplementary Figure 2 for a 2-dimensional embedding. The upper planes (in green)
show a predictee (red triangle) and the library vectors (blue squares). The lower surfaces
illustrate the weighting functions for a linear map (Supplementary Figure 2a) and a
nonlinear map (Supplementary Figure 2b). The time series reflects nonlinear dynamics
when nonlinear mappings (Supplementary Figure 2b) outperform the corresponding
linear map (Supplementary Figure 2b) in out-of-sample forecasts.
Supplementary Figure 1. An example illustrating simplex projection for a time series
embedded in a two dimensional (E = 2) lagged-coordinate space. The figure shows a
one-step forward forecast using nearby neighbors from the library. The predictee Yt is a
two dimensional vector formed from points in the prediction set. The three (E+1) nearest
neighbors () from the library set that form the domain simplex are projected one step
forward to yield the range simplex (). The forecast (

ˆYt 1) is a weighted average of the
values given by the range simplex () with weights depending exponentially on
proximity of the neighboring domain vectors () to the predictee Yt.
Supplementary Figure 2. Examples illustrating the S-map procedure for a linear map (a)
and a nonlinear map (b) with embedding dimension equal to two. The upper plane in
each frame represents the lag coordinate embedding of the library file where E = 2. The
lower panel is a geometric representation of the weighting function

(d)  edit /d
, where
 = 0 (left panel) and  > 0 (right panel). In the linear map, all library vectors (blue
squares) are weighted equally. In the nonlinear map, the points closest to the predictee
(red triangle) contribute most heavily to the forecast.
References
1. Sugihara, G. & May, R. M. Nonlinear forecasting as a way of distinguishing chaos
from measurement error in time-series. Nature 344, 734-741 (1990).
2. Sugihara, G. Nonlinear forecasting for the classification of natural time-series. Philos.
T. Roy. Soc. A 348, 477-495 (1994).
3. Crutchfield, J. P. Prediction and stability in classical mechanics. Bachelor’s Thesis,
University of California, Santa Cruz (1979).
4. Takens, F. Detecting strange attractors in turbulence. Lect. Notes Math. 898, 366-381
(1981).
5. Farmer, J. D. & Sidorowich, J. J. in Evolution, learning and cognition (ed. Lee., Y. C.)
277-304 (World Scientific Press, New York, 1989).

Weitere ähnliche Inhalte

Was ist angesagt?

Scaling Transform Methods For Compressing a 2D Graphical image
Scaling Transform Methods For Compressing a 2D Graphical image Scaling Transform Methods For Compressing a 2D Graphical image
Scaling Transform Methods For Compressing a 2D Graphical image acijjournal
 
EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...
EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...
EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...ijfcstjournal
 
case study of curve fitting
case study of curve fittingcase study of curve fitting
case study of curve fittingAdarsh Patel
 
Size measurement and estimation
Size measurement and estimationSize measurement and estimation
Size measurement and estimationLouis A. Poulin
 
Spatial data mining
Spatial data miningSpatial data mining
Spatial data miningMITS Gwalior
 
QTML2021 UAP Quantum Feature Map
QTML2021 UAP Quantum Feature MapQTML2021 UAP Quantum Feature Map
QTML2021 UAP Quantum Feature MapHa Phuong
 
"FingerPrint Recognition Using Principle Component Analysis(PCA)”
"FingerPrint Recognition Using Principle Component Analysis(PCA)”"FingerPrint Recognition Using Principle Component Analysis(PCA)”
"FingerPrint Recognition Using Principle Component Analysis(PCA)”Er. Arpit Sharma
 
Use of Eigenvector Centrality to Detect Graph Isomorphism
Use of Eigenvector Centrality to Detect Graph IsomorphismUse of Eigenvector Centrality to Detect Graph Isomorphism
Use of Eigenvector Centrality to Detect Graph Isomorphismcsandit
 
Mar 8 single_map_analysis_1
Mar 8 single_map_analysis_1Mar 8 single_map_analysis_1
Mar 8 single_map_analysis_1dellissimo
 
Geographical information system unit 5
Geographical information  system unit 5Geographical information  system unit 5
Geographical information system unit 5WE-IT TUTORIALS
 
Extrapolation
ExtrapolationExtrapolation
Extrapolationjonathan
 
Some Engg. Applications of Matrices and Partial Derivatives
Some Engg. Applications of Matrices and Partial DerivativesSome Engg. Applications of Matrices and Partial Derivatives
Some Engg. Applications of Matrices and Partial DerivativesSanjaySingh011996
 
Models of spatial process by sushant
Models of spatial process by sushantModels of spatial process by sushant
Models of spatial process by sushantsushantsawant13
 
Advanced Support Vector Machine for classification in Neural Network
Advanced Support Vector Machine for classification  in Neural NetworkAdvanced Support Vector Machine for classification  in Neural Network
Advanced Support Vector Machine for classification in Neural NetworkAshwani Jha
 
4.2 spatial data mining
4.2 spatial data mining4.2 spatial data mining
4.2 spatial data miningKrish_ver2
 

Was ist angesagt? (20)

Scaling Transform Methods For Compressing a 2D Graphical image
Scaling Transform Methods For Compressing a 2D Graphical image Scaling Transform Methods For Compressing a 2D Graphical image
Scaling Transform Methods For Compressing a 2D Graphical image
 
EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...
EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...
EXPLOITING THE DISCRIMINATING POWER OF THE EIGENVECTOR CENTRALITY MEASURE TO ...
 
case study of curve fitting
case study of curve fittingcase study of curve fitting
case study of curve fitting
 
Size measurement and estimation
Size measurement and estimationSize measurement and estimation
Size measurement and estimation
 
Spatial data mining
Spatial data miningSpatial data mining
Spatial data mining
 
QTML2021 UAP Quantum Feature Map
QTML2021 UAP Quantum Feature MapQTML2021 UAP Quantum Feature Map
QTML2021 UAP Quantum Feature Map
 
"FingerPrint Recognition Using Principle Component Analysis(PCA)”
"FingerPrint Recognition Using Principle Component Analysis(PCA)”"FingerPrint Recognition Using Principle Component Analysis(PCA)”
"FingerPrint Recognition Using Principle Component Analysis(PCA)”
 
Use of Eigenvector Centrality to Detect Graph Isomorphism
Use of Eigenvector Centrality to Detect Graph IsomorphismUse of Eigenvector Centrality to Detect Graph Isomorphism
Use of Eigenvector Centrality to Detect Graph Isomorphism
 
Mar 8 single_map_analysis_1
Mar 8 single_map_analysis_1Mar 8 single_map_analysis_1
Mar 8 single_map_analysis_1
 
Geographical information system unit 5
Geographical information  system unit 5Geographical information  system unit 5
Geographical information system unit 5
 
Mathematical modeling
Mathematical modelingMathematical modeling
Mathematical modeling
 
Extrapolation
ExtrapolationExtrapolation
Extrapolation
 
Extrapolation
ExtrapolationExtrapolation
Extrapolation
 
Some Engg. Applications of Matrices and Partial Derivatives
Some Engg. Applications of Matrices and Partial DerivativesSome Engg. Applications of Matrices and Partial Derivatives
Some Engg. Applications of Matrices and Partial Derivatives
 
Models of spatial process by sushant
Models of spatial process by sushantModels of spatial process by sushant
Models of spatial process by sushant
 
Pca ppt
Pca pptPca ppt
Pca ppt
 
Isam2_v1_2
Isam2_v1_2Isam2_v1_2
Isam2_v1_2
 
Pca ankita dubey
Pca ankita dubeyPca ankita dubey
Pca ankita dubey
 
Advanced Support Vector Machine for classification in Neural Network
Advanced Support Vector Machine for classification  in Neural NetworkAdvanced Support Vector Machine for classification  in Neural Network
Advanced Support Vector Machine for classification in Neural Network
 
4.2 spatial data mining
4.2 spatial data mining4.2 spatial data mining
4.2 spatial data mining
 

Andere mochten auch

Andere mochten auch (14)

Vidyasagar rocond09
Vidyasagar rocond09Vidyasagar rocond09
Vidyasagar rocond09
 
Efficient projections
Efficient projectionsEfficient projections
Efficient projections
 
Repeat after me_8_14_13
Repeat after me_8_14_13Repeat after me_8_14_13
Repeat after me_8_14_13
 
Comparison
ComparisonComparison
Comparison
 
1 s2.0-s0167268105000983-main
1 s2.0-s0167268105000983-main1 s2.0-s0167268105000983-main
1 s2.0-s0167268105000983-main
 
C plus plus_for_quantitative_finance
C plus plus_for_quantitative_financeC plus plus_for_quantitative_finance
C plus plus_for_quantitative_finance
 
Codeconventions 150003
Codeconventions 150003Codeconventions 150003
Codeconventions 150003
 
G
GG
G
 
Jens valloe christiansen
Jens valloe christiansenJens valloe christiansen
Jens valloe christiansen
 
4587 11094-1-pb
4587 11094-1-pb4587 11094-1-pb
4587 11094-1-pb
 
Repeat after me_8_14_13
Repeat after me_8_14_13Repeat after me_8_14_13
Repeat after me_8_14_13
 
Mvm
MvmMvm
Mvm
 
Programowanie automatycznych strategii transakcyjnych czesc podstawowa
Programowanie automatycznych strategii transakcyjnych   czesc podstawowaProgramowanie automatycznych strategii transakcyjnych   czesc podstawowa
Programowanie automatycznych strategii transakcyjnych czesc podstawowa
 
trading_automatyczny-waszczyk-probka
trading_automatyczny-waszczyk-probkatrading_automatyczny-waszczyk-probka
trading_automatyczny-waszczyk-probka
 

Ähnlich wie Hsieh etal spl

Performance Improvement of Vector Quantization with Bit-parallelism Hardware
Performance Improvement of Vector Quantization with Bit-parallelism HardwarePerformance Improvement of Vector Quantization with Bit-parallelism Hardware
Performance Improvement of Vector Quantization with Bit-parallelism HardwareCSCJournals
 
USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM
USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM
USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM cscpconf
 
Computer graphics notes 2 tutorials duniya
Computer graphics notes 2   tutorials duniyaComputer graphics notes 2   tutorials duniya
Computer graphics notes 2 tutorials duniyaTutorialsDuniya.com
 
Drobics, m. 2001: datamining using synergiesbetween self-organising maps and...
Drobics, m. 2001:  datamining using synergiesbetween self-organising maps and...Drobics, m. 2001:  datamining using synergiesbetween self-organising maps and...
Drobics, m. 2001: datamining using synergiesbetween self-organising maps and...ArchiLab 7
 
Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...
Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...
Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...Cemal Ardil
 
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...csandit
 
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...cscpconf
 
Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...
Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...
Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...IOSRJECE
 
Data-Driven Motion Estimation With Spatial Adaptation
Data-Driven Motion Estimation With Spatial AdaptationData-Driven Motion Estimation With Spatial Adaptation
Data-Driven Motion Estimation With Spatial AdaptationCSCJournals
 
Interpolation of-geofield-parameters
Interpolation of-geofield-parametersInterpolation of-geofield-parameters
Interpolation of-geofield-parametersCemal Ardil
 
imageCorrectionLinearDiffusion
imageCorrectionLinearDiffusionimageCorrectionLinearDiffusion
imageCorrectionLinearDiffusionKellen Betts
 
Image Processing
Image ProcessingImage Processing
Image ProcessingTuyen Pham
 
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...ieijjournal
 
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...ieijjournal
 
A Rapid Location Independent Full Tensor Gravity Algorithm
A Rapid Location Independent Full Tensor Gravity AlgorithmA Rapid Location Independent Full Tensor Gravity Algorithm
A Rapid Location Independent Full Tensor Gravity AlgorithmPioneer Natural Resources
 

Ähnlich wie Hsieh etal spl (20)

Performance Improvement of Vector Quantization with Bit-parallelism Hardware
Performance Improvement of Vector Quantization with Bit-parallelism HardwarePerformance Improvement of Vector Quantization with Bit-parallelism Hardware
Performance Improvement of Vector Quantization with Bit-parallelism Hardware
 
Diffusion Homework Help
Diffusion Homework HelpDiffusion Homework Help
Diffusion Homework Help
 
USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM
USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM
USE OF EIGENVECTOR CENTRALITY TO DETECT GRAPH ISOMORPHISM
 
icarsn
icarsnicarsn
icarsn
 
Computer graphics notes 2 tutorials duniya
Computer graphics notes 2   tutorials duniyaComputer graphics notes 2   tutorials duniya
Computer graphics notes 2 tutorials duniya
 
Dycops2019
Dycops2019 Dycops2019
Dycops2019
 
Drobics, m. 2001: datamining using synergiesbetween self-organising maps and...
Drobics, m. 2001:  datamining using synergiesbetween self-organising maps and...Drobics, m. 2001:  datamining using synergiesbetween self-organising maps and...
Drobics, m. 2001: datamining using synergiesbetween self-organising maps and...
 
BNL_Research_Report
BNL_Research_ReportBNL_Research_Report
BNL_Research_Report
 
Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...
Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...
Neuro -fuzzy-networks-for-identification-of-mathematical-model-parameters-of-...
 
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
 
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
STATE SPACE GENERATION FRAMEWORK BASED ON BINARY DECISION DIAGRAM FOR DISTRIB...
 
Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...
Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...
Investigation on the Pattern Synthesis of Subarray Weights for Low EMI Applic...
 
Data-Driven Motion Estimation With Spatial Adaptation
Data-Driven Motion Estimation With Spatial AdaptationData-Driven Motion Estimation With Spatial Adaptation
Data-Driven Motion Estimation With Spatial Adaptation
 
Interpolation of-geofield-parameters
Interpolation of-geofield-parametersInterpolation of-geofield-parameters
Interpolation of-geofield-parameters
 
imageCorrectionLinearDiffusion
imageCorrectionLinearDiffusionimageCorrectionLinearDiffusion
imageCorrectionLinearDiffusion
 
Image Processing
Image ProcessingImage Processing
Image Processing
 
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
 
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
FITTED OPERATOR FINITE DIFFERENCE METHOD FOR SINGULARLY PERTURBED PARABOLIC C...
 
Kriging
KrigingKriging
Kriging
 
A Rapid Location Independent Full Tensor Gravity Algorithm
A Rapid Location Independent Full Tensor Gravity AlgorithmA Rapid Location Independent Full Tensor Gravity Algorithm
A Rapid Location Independent Full Tensor Gravity Algorithm
 

Mehr von Tomasz Waszczyk

Wzory portfel-inwestycyjnyrynek-kapitalowy
Wzory portfel-inwestycyjnyrynek-kapitalowyWzory portfel-inwestycyjnyrynek-kapitalowy
Wzory portfel-inwestycyjnyrynek-kapitalowyTomasz Waszczyk
 
2012 11-03-boersenfuehrerschein-teil6-technische-analyse
2012 11-03-boersenfuehrerschein-teil6-technische-analyse2012 11-03-boersenfuehrerschein-teil6-technische-analyse
2012 11-03-boersenfuehrerschein-teil6-technische-analyseTomasz Waszczyk
 
Quant congressusa2011algotradinglast
Quant congressusa2011algotradinglastQuant congressusa2011algotradinglast
Quant congressusa2011algotradinglastTomasz Waszczyk
 
Sdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skriptSdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skriptTomasz Waszczyk
 
In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...
In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...
In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...Tomasz Waszczyk
 
Sdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skriptSdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skriptTomasz Waszczyk
 
8 analyse der_aktienentwicklung
8 analyse der_aktienentwicklung8 analyse der_aktienentwicklung
8 analyse der_aktienentwicklungTomasz Waszczyk
 
Open splicedds espercep-webinar
Open splicedds espercep-webinarOpen splicedds espercep-webinar
Open splicedds espercep-webinarTomasz Waszczyk
 
Zadania z programowania c
Zadania z programowania c  Zadania z programowania c
Zadania z programowania c Tomasz Waszczyk
 
Tadeusz czachórski modele kolejkowe w ocenie efektywności sieci i systemów ...
Tadeusz czachórski   modele kolejkowe w ocenie efektywności sieci i systemów ...Tadeusz czachórski   modele kolejkowe w ocenie efektywności sieci i systemów ...
Tadeusz czachórski modele kolejkowe w ocenie efektywności sieci i systemów ...Tomasz Waszczyk
 

Mehr von Tomasz Waszczyk (17)

Wzory portfel-inwestycyjnyrynek-kapitalowy
Wzory portfel-inwestycyjnyrynek-kapitalowyWzory portfel-inwestycyjnyrynek-kapitalowy
Wzory portfel-inwestycyjnyrynek-kapitalowy
 
2012 11-03-boersenfuehrerschein-teil6-technische-analyse
2012 11-03-boersenfuehrerschein-teil6-technische-analyse2012 11-03-boersenfuehrerschein-teil6-technische-analyse
2012 11-03-boersenfuehrerschein-teil6-technische-analyse
 
Lm mme090913
Lm mme090913Lm mme090913
Lm mme090913
 
Quant congressusa2011algotradinglast
Quant congressusa2011algotradinglastQuant congressusa2011algotradinglast
Quant congressusa2011algotradinglast
 
1108.1170
1108.11701108.1170
1108.1170
 
Efficient projections
Efficient projectionsEfficient projections
Efficient projections
 
Sdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skriptSdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skript
 
In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...
In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...
In tech an-innovative_systematic_approach_to_financial_portfolio_management_v...
 
Sdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skriptSdv 0405 design-pattern_thc_jps_skript
Sdv 0405 design-pattern_thc_jps_skript
 
Div dax d
Div dax dDiv dax d
Div dax d
 
8 analyse der_aktienentwicklung
8 analyse der_aktienentwicklung8 analyse der_aktienentwicklung
8 analyse der_aktienentwicklung
 
2013 03
2013 032013 03
2013 03
 
Open splicedds espercep-webinar
Open splicedds espercep-webinarOpen splicedds espercep-webinar
Open splicedds espercep-webinar
 
Trailingstop
TrailingstopTrailingstop
Trailingstop
 
Elsholz stoll js_03_10
Elsholz stoll js_03_10Elsholz stoll js_03_10
Elsholz stoll js_03_10
 
Zadania z programowania c
Zadania z programowania c  Zadania z programowania c
Zadania z programowania c
 
Tadeusz czachórski modele kolejkowe w ocenie efektywności sieci i systemów ...
Tadeusz czachórski   modele kolejkowe w ocenie efektywności sieci i systemów ...Tadeusz czachórski   modele kolejkowe w ocenie efektywności sieci i systemów ...
Tadeusz czachórski modele kolejkowe w ocenie efektywności sieci i systemów ...
 

Kürzlich hochgeladen

Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsMaria Levchenko
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Drew Madelung
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processorsdebabhi2
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfsudhanshuwaghmare1
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Miguel Araújo
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Servicegiselly40
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CVKhem
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking MenDelhi Call girls
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024The Digital Insurer
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Scriptwesley chun
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsJoaquim Jorge
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEarley Information Science
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreternaman860154
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityPrincipled Technologies
 

Kürzlich hochgeladen (20)

Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdf
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Service
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CV
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Script
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivity
 

Hsieh etal spl

  • 1. Supplementary Methods To determine whether a time series reflects linear or nonlinear processes we compare the out-of-sample forecast skill of a linear model versus an equivalent nonlinear model. To do this, we apply a two-step procedure: 1) we use simplex-projection1 to identify the best embedding dimension, and 2) we use this embedding in the S-map procedure2 to assess the nonlinearity of the time series. In both cases, model performance is evaluated out-of-sample with the time series divided into equal halves. The first half (library set, X) is used to build the model, while the second half (prediction set, Y) is reserved to judge the out-of-sample performance of model forecasts originating from the library. This forecast protocol is a rigorous standard that avoids model over-fitting or arbitrary fits to the data. Simplex projection Simplex projection is a nearest-neighbor forecasting algorithm that involves tracking the forward evolution of nearby points in an embedding (a lagged coordinate state space reconstruction). 3, 4, 5 Thus, similar past events are used to forecast the future, with the important caveat that the dimensionality of the embedding determines what past events are similar (nearby) to the predictee. Given a library set of n points, we generate an E-dimensional embedding from the time series data using lagged coordinates so that the library vectors Xt , {xt , xt-, xt-2 … xt-(E-1)} and prediction vectors Yt , {yt , yt-, yt-2 … yt-(E-1)} are points in E-dimensional state space (where =1, and t  [1, 2, … , n] ). We then choose the E+1 neighbors of Yt from the library set X that form the vertices of the smallest simplex that contains Yt as an interior point. The forecast is based on how these nearby library points (domain simplex) move forward in time (range simplex). Again, the key is in the
  • 2. definition of “nearby” which depends critically on the dimension of the embedding. Thus, the domain simplex is projected forward into its range by incrementing the time index of each neighbor in the domain simplex by 1. The forecast is simply a weighted average of the values given by the range simplex. The weights depend exponentially on proximity of the library vectors in the domain simplex to the predictee vector Yt (illustrated in Supplementary Figure 1). For this study an exploratory series of embedding dimensions (E) ranging from 1 to 20 (or higher) are used to evaluate the prediction, and the best E is chosen based on prediction skill. This embedding is then used in the S-map procedure. S-map algorithm S-maps are an extension of standard linear autoregressive models in which the coefficients depend on the location of the predictee Yt in an E-dimensional embedding. New coefficients are recalculated (from the library set X) by singular value decomposition (SVD) for each new prediction. In this calculation, the weight given to each vector in the library depends on how close that vector Xt is to the predictee Yt. The extent of this weighting is determined by the parameter . As above, we generate an E-dimensional embedding from points in the library using lagged coordinates to obtain an embedded time series with vectors Xt  RE+1 , where Xt(0) = 1 is the constant term in the solution of Eq. (S2) below. Let the time series observation in the prediction set Tp time steps forward be Yt+Tp(1)=Y(t). Then the forecast for Y(t) is  ˆYt  Ct ( j)Xt ( j) j 0 E  (S1) For our analysis, we chose TP = 1. For each E-dimensional predictee vector Yt, C is solved by SVD using the library set as follows: B = AC, (S2)
  • 3. where  Bi (|| Xi Yt ||)Yi,  Aij (|| Xi Yt ||)Xi( j), and  (d)  edit /d    0, dit is the distance between Yt and the ith neighbor vector Xi in the library embedding, and the scale vector,  d , is the average distance between neighbors in the library. Note that A has dimensions  n (E 1), where  n = size of the library. Again, a different map is generated for each forecast, with the weightings in each map depending on the location of the predictee in the E-dimensional state space. This weighting procedure is governed by the tuning parameter , where  = 0 gives a global linear map, and increasing values of  give increasingly local or nonlinear mappings. Note that when  = 0, all vectors are weighted equally so a single (global) linear map can be used for all predictions. In the case where  > 0, vectors closest to the predictee in state-space are weighted more heavily in the SVD solution. Such forecasts emphasize local information in the library set, and are therefore nonlinear. The method is illustrated in the Supplementary Figure 2 for a 2-dimensional embedding. The upper planes (in green) show a predictee (red triangle) and the library vectors (blue squares). The lower surfaces illustrate the weighting functions for a linear map (Supplementary Figure 2a) and a nonlinear map (Supplementary Figure 2b). The time series reflects nonlinear dynamics when nonlinear mappings (Supplementary Figure 2b) outperform the corresponding linear map (Supplementary Figure 2b) in out-of-sample forecasts.
  • 4. Supplementary Figure 1. An example illustrating simplex projection for a time series embedded in a two dimensional (E = 2) lagged-coordinate space. The figure shows a one-step forward forecast using nearby neighbors from the library. The predictee Yt is a two dimensional vector formed from points in the prediction set. The three (E+1) nearest neighbors () from the library set that form the domain simplex are projected one step forward to yield the range simplex (). The forecast (  ˆYt 1) is a weighted average of the values given by the range simplex () with weights depending exponentially on proximity of the neighboring domain vectors () to the predictee Yt.
  • 5. Supplementary Figure 2. Examples illustrating the S-map procedure for a linear map (a) and a nonlinear map (b) with embedding dimension equal to two. The upper plane in each frame represents the lag coordinate embedding of the library file where E = 2. The lower panel is a geometric representation of the weighting function  (d)  edit /d , where  = 0 (left panel) and  > 0 (right panel). In the linear map, all library vectors (blue squares) are weighted equally. In the nonlinear map, the points closest to the predictee (red triangle) contribute most heavily to the forecast. References 1. Sugihara, G. & May, R. M. Nonlinear forecasting as a way of distinguishing chaos from measurement error in time-series. Nature 344, 734-741 (1990). 2. Sugihara, G. Nonlinear forecasting for the classification of natural time-series. Philos. T. Roy. Soc. A 348, 477-495 (1994). 3. Crutchfield, J. P. Prediction and stability in classical mechanics. Bachelor’s Thesis, University of California, Santa Cruz (1979).
  • 6. 4. Takens, F. Detecting strange attractors in turbulence. Lect. Notes Math. 898, 366-381 (1981). 5. Farmer, J. D. & Sidorowich, J. J. in Evolution, learning and cognition (ed. Lee., Y. C.) 277-304 (World Scientific Press, New York, 1989).