1 Recurrence analysis and Fisheries Fisheries as a complex systems Traditional science operates on the assumption that natural systems like fish populations.

Slides:



Advertisements
Similar presentations
Spatial point patterns and Geostatistics an introduction
Advertisements

11-1 Empirical Models Many problems in engineering and science involve exploring the relationships between two or more variables. Regression analysis.
Computational Statistics. Basic ideas  Predict values that are hard to measure irl, by using co-variables (other properties from the same measurement.
Principal Component Analysis Based on L1-Norm Maximization Nojun Kwak IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008.
From
Inference for Regression
FTP Biostatistics II Model parameter estimations: Confronting models with measurements.
AGC DSP AGC DSP Professor A G Constantinides©1 Modern Spectral Estimation Modern Spectral Estimation is based on a priori assumptions on the manner, the.
G. Alonso, D. Kossmann Systems Group
STAT 497 APPLIED TIME SERIES ANALYSIS
Mining for High Complexity Regions Using Entropy and Box Counting Dimension Quad-Trees Rosanne Vetro, Wei Ding, Dan A. Simovici Computer Science Department.
Copyright © Cengage Learning. All rights reserved. 9 Inferences Based on Two Samples.
Chapter 10 Simple Regression.
Data Sources The most sophisticated forecasting model will fail if it is applied to unreliable data Data should be reliable and accurate Data should be.
Evaluating Hypotheses
2008 Chingchun 1 Bootstrap Chingchun Huang ( 黃敬群 ) Vision Lab, NCTU.
Discovering and Describing Relationships
Atul Singh Junior Undergraduate CSE, IIT Kanpur.  Dimension reduction is a technique which is used to represent a high dimensional data in a more compact.
Experimental Evaluation
Tracking with Linear Dynamic Models. Introduction Tracking is the problem of generating an inference about the motion of an object given a sequence of.
Separate multivariate observations
©2003/04 Alessandro Bogliolo Background Information theory Probability theory Algorithms.
Statistical Methods For Engineers ChE 477 (UO Lab) Larry Baxter & Stan Harding Brigham Young University.
1 CE 530 Molecular Simulation Lecture 7 David A. Kofke Department of Chemical Engineering SUNY Buffalo
1 Statistical Mechanics and Multi- Scale Simulation Methods ChBE Prof. C. Heath Turner Lecture 11 Some materials adapted from Prof. Keith E. Gubbins:
Fundamentals of Data Analysis Lecture 4 Testing of statistical hypotheses.
Statistical Analysis A Quick Overview. The Scientific Method Establishing a hypothesis (idea) Collecting evidence (often in the form of numerical data)
Reconstructed Phase Space (RPS)
8 th Grade Math Common Core Standards. The Number System 8.NS Know that there are numbers that are not rational, and approximate them by rational numbers.
Time Series Data Analysis - I Yaji Sripada. Dept. of Computing Science, University of Aberdeen2 In this lecture you learn What are Time Series? How to.
The Examination of Residuals. Examination of Residuals The fitting of models to data is done using an iterative approach. The first step is to fit a simple.
Introduction to Analytical Chemistry
Applications of Neural Networks in Time-Series Analysis Adam Maus Computer Science Department Mentor: Doctor Sprott Physics Department.
Progress in identification of damping: Energy-based method with incomplete and noisy data Marco Prandina University of Liverpool.
TODAY we will Review what we have learned so far about Regression Develop the ability to use Residual Analysis to assess if a model (LSRL) is appropriate.
Statistical Methods II&III: Confidence Intervals ChE 477 (UO Lab) Lecture 5 Larry Baxter, William Hecker, & Ron Terry Brigham Young University.
Time series Model assessment. Tourist arrivals to NZ Period is quarterly.
STATISTICAL COMPLEXITY ANALYSIS Dr. Dmitry Nerukh Giorgos Karvounis.
© Copyright McGraw-Hill Correlation and Regression CHAPTER 10.
1 Chapter 9 Detection of Spread-Spectrum Signals.
Some figures adapted from a 2004 Lecture by Larry Liebovitch, Ph.D. Chaos BIOL/CMSC 361: Emergence 1/29/08.
Introduction: Brain Dynamics Jaeseung Jeong, Ph.D Department of Bio and Brain Engineering, KAIST.
CCN COMPLEX COMPUTING NETWORKS1 This research has been supported in part by European Commission FP6 IYTE-Wireless Project (Contract No: )
CHAPTER 5 SIGNAL SPACE ANALYSIS
ECE-7000: Nonlinear Dynamical Systems Overfitting and model costs Overfitting  The more free parameters a model has, the better it can be adapted.
Experimentation in Computer Science (Part 2). Experimentation in Software Engineering --- Outline  Empirical Strategies  Measurement  Experiment Process.
Chapter 8: Simple Linear Regression Yang Zhenlin.
Chapter 20 Classification and Estimation Classification – Feature selection Good feature have four characteristics: –Discrimination. Features.
1 Challenge the future Chaotic Invariants for Human Action Recognition Ali, Basharat, & Shah, ICCV 2007.
McGraw-Hill/IrwinCopyright © 2009 by The McGraw-Hill Companies, Inc. All Rights Reserved. Simple Linear Regression Analysis Chapter 13.
Spectrum Sensing In Cognitive Radio Networks
1 EE571 PART 3 Random Processes Huseyin Bilgekul Eeng571 Probability and astochastic Processes Department of Electrical and Electronic Engineering Eastern.
Learning Chaotic Dynamics from Time Series Data A Recurrent Support Vector Machine Approach Vinay Varadan.
September 28, 2000 Improved Simultaneous Data Reconciliation, Bias Detection and Identification Using Mixed Integer Optimization Methods Presented by:
ECE-7000: Nonlinear Dynamical Systems 3. Phase Space Methods 3.1 Determinism: Uniqueness in phase space We Assume that the system is linear stochastic.
[Chaos in the Brain] Nonlinear dynamical analysis for neural signals Jaeseung Jeong, Ph.D Department of Bio and Brain Engineering, KAIST.
Page 0 of 5 Dynamical Invariants of an Attractor and potential applications for speech data Saurabh Prasad Intelligent Electronic Systems Human and Systems.
Fundamentals of Data Analysis Lecture 10 Correlation and regression.
ANALYSES OF THE CHAOTIC BEHAVIOR OF THE ELECTRICITY PRICE SERIES
Multiple Random Variables and Joint Distributions
Regression Analysis AGEC 784.
Why Stochastic Hydrology ?
Dimension Review Many of the geometric structures generated by chaotic map or differential dynamic systems are extremely complex. Fractal : hard to define.
Nonlinear Structure in Regression Residuals
Application of Independent Component Analysis (ICA) to Beam Diagnosis
Correlation and Regression
Autocorrelation.
Discrete Event Simulation - 4
Lecture # 2 MATHEMATICAL STATISTICS
Autocorrelation.
Presentation transcript:

1 Recurrence analysis and Fisheries Fisheries as a complex systems Traditional science operates on the assumption that natural systems like fish populations exist in balance. But in reality they are in constant flux - spiking up and down, with many overlapping cycles - and can be upset by even tiny changes. Now there are a growing emphasis both on interdependencies in large marine ecosystems and on the non-linear or chaotic changes that are prominent features of these systems.

2 Recurrence analysis and Fisheries Time series analysis The main aims when investigate a nonstionarity time series are Characterisation. When a dynamics underlying a non stationarity of time series is similar at beginning and end but not at middle, the aim of techniques which analyse the non stationarity is to select this information in convenient form. Prediction. An accurate predictions (Priestley 1988, Stark 1993,Weigend et alt. 1995) for a non stationariety time series can be obtained modifying prediction algorithms. This procedures is strongly connected to characterisation and modelling time series.

3 Recurrence analysis and Fisheries Time series analysis Change point detection. The dynamics which characterises a time series can have a single point of change. The objective of time series analysis is to identify this change point (Lima 1995, Kennel 1996). Hypothesis testing. In some case there may be reason to believe that a time series is stationary. The objective is to develop hypothesis tests for the null hypothesis. If the null hypothesis is accepted one can then have confidence in applying techniques of stationarity time series analysis. The recurrence plots provide an unified framework for addressing all those objectives

4 Recurrence analysis and Fisheries Recurrence plots approach The recurrence plots are a visualisation tool for analysing experimental data. The basic feature of this tool is to reveal correlation in the data that are not easily detected in the original time series. It not requires any assumptions on the stationarity of time series, then Rps are particularly used in the analysis of systems whose dynamics may be changing. Some author (Zbilut, Webber, giuliani, Trulla) have highlighted that RPs can be compared to classical approaches for analysing chaotic data, especially in its ability to detect bifurcation. The fundamental assumption underlying the idea of the recurrence plots is that an observable time series (a sequence of observations) is the realization of some dynamical process, the interaction of the relevant variables over time

5 Recurrence analysis and Fisheries Takens theorem This theorem says that it can recreate a topologically equivalent picture of the original multidimensional system behavior by using the time series of a single observable variable. The basic idea is that the effect of all the other (unobserved) variables is already reflected in the series of the observed output. Furthermore, the rules that govern the behavior of the original system can be recovered from its output.

6 Recurrence analysis and Fisheries Recurrence plots approach The recurrence plots basis on the reconstruction space state. The reconstruction space state is the first steps to analyse a time series in a context of dynamical systems theory when the only information is contained in a time series. This approach is founded on flow of information from unobserved variables to observed variables “past and future of a time series contain information about unobserved state variables that can be used to define a state at the present time” (M. Casdagli, S. Eubank. J. D. Farmer, J. Gibson, 1991). Although the theoretical approach of Takens’ theorem highlight that the choice of coordinate which representing an embedding is indifferent, in practical analysis it is demonstrated as the choice of coordinate affects the predictions result.

7 Recurrence analysis and Fisheries Methods for reconstruction space state The methods for reconstruction space state are: Delay coordinates Derivative coordinates Global principal value decomposition Although all three methods are applied for the reconstruction space state, the delay coordinates is most used (VRA basis on this method).

8 Recurrence analysis and Fisheries Delayed coordinate In VRA, a one-dimensional time series from a data file is expanded into a higher- dimensional space, in which the dynamic of the underlying generator takes place. The delayed coordinate embedding recreates a phase space portrait of the dynamical system under study from a single (scalar) time series. To expand a one-dimensional signal into an M-dimensional phase space, one substitutes each observation in the original signal X(t) with vector y(i) = {x(i), x(i - d), x(i - 2d), …, x(i - (m-1)d}, i is the time index, m is the embedding dimension d is the time delay. As a result, we have a series of vectors: Y = {y(1), y(2), y(3), …, y(N-(m-1)d)}, N is the length of the original series. The idea of such reconstruction is to capture the original system states at each time we have an observation of that system output. Each unknown state S(t) at time t is approximated by a vector of delayed coordinates Y(t) = { x(t), x(t - d), x(t - 2d), …, x(t - (m-1)d }

9 Recurrence analysis and Fisheries Spatio-Temporal Entropy Spatio-Temporal Entropy (STE) measures the image "structureness" in both space and time domains. Essentially, it compares the global distribution of colors over the entire recurrence plot with the distribution of colors over each diagonal line of the recurrence plot. The higher the combined differences between the global distribution and the distributions over the individual diagonal lines, the more structured the image is. In physical terms, this quantity compares the distribution of distances between all pairs of vectors in the reconstructed state space with that of distances between different orbits evolving in time. The result is normalized and presented as a percentage of "maximum" entropy (randomness). That is, 100% entropy means the absence of any structure whatsoever (uniform distribution of colors, pure randomness), while 0% entropy implies "perfect" structure (distinct color patterns, perfect "structureness" and predictability). Thus, the following range of spatio-temporal entropy should be expected for different signals. Notice that the closer are the values of the embedding dimension and the time delay to the “true”values, the lower is the entropy of the recurrence plot

10 Recurrence analysis and Fisheries Embedding dimension The analytical methods for estimating the embedding dimension is the false nearest neighbours method. The False Nearest Neighbors is a method of choosing the minimum embedding dimension of a one-dimensional time series, suggested by Kennel et al. This method finds the nearest neighbor of every point in a given dimension, then checks to see if these points are still close neighbors in one higher dimension. The percentage of False Nearest Neighbors should drop to zero when the appropriate embedding dimension has been reached.

11 Recurrence analysis and Fisheries The False Nearest Neighbors

12 Recurrence analysis and Fisheries The False Nearest Neighbors Ideally, d should be large enough to unfold the system trajectories from self- overlaps, but not too large, the noise will amplify. The rule of thumb is to set m to m<= 2N+1, where N is the number of operating variables, or degrees of freedom, in the dynamical system under study

13 Recurrence analysis and Fisheries Delay time The analytical methods for estimating the delay time is the mutual information function. Mutual information is a general measure, based on information theory, of the extent to which the values in a time series can be predicted by earlier values. It is not limited to linear dependence as is the autocorrelation function. It measures the state predictability or the memory of a system.

14 Recurrence analysis and Fisheries Mutual information function Mutual information function can be used to determine the “optimal” value of the time delay for the state space reconstruction. The idea is that a good choice for the time delay T is one that, given the state of the system X(t), provides maximum new information with measurement at X(t+T). Mutual information is the answer to the question, "Given a measurement of X(t), how many bits on the average can be predicted about X(t+T)?" A graph of I(T) starts off very high (given a measurement X(t), we know as many bits as possible about X(t+0)=X(t)). As T is increased, I(T) decreases, then usually rises again. It is suggested that the value of time delay where I(T) reaches its first minimum be used for the state space reconstruction.

15 Recurrence analysis and Fisheries Mutual information function an example The mutual information I(k) (Farmer1982 ) is a generalization of the correlation function. It measures the state predictability or the memory of a system, represented by a sequence of certain symbols. In the following we consider sequences in which only two different symbols (say 0 and 1) can occur. Then, I(k) defined by quantifies the average dependence of two symbols over k time steps, where p i is the probability of the symbol of a symbol string, and is the joint probability that the symbol and steps k later the symbol occurs. In particular, peaks in I(k) exhibit the levels of memory quantitatively. This description is especially appropriate for nonlinear systems because the joint probabilities reflect more general dependencies within a symbolic string than the autocorrelation function.

16 Recurrence analysis and Fisheries Mutual information function an example (continued) In the case of white noise, mutual information I(k) vanishes for all,reflecting that white noise is a process without any memory. Periodic processes are characterized by peaks at the multiples of the period in k. Chaotic regimes yield a decrease of mutual information I(k) with growing k (Hempelmann and Kurths 1990). Note that structures of particular interest here are detected by mutual information but not by autocorrelation function. For short sequences, artifacts may occur in the calculation of I(k). Therefore, we propose a method to judge the statistical significance of peaks in I(k) by a statistical, semi-empiric method, called randomization (e.g. Random 1990).

17 Recurrence analysis and Fisheries Application fisheries data Delay 1 Dimension 1

18 Recurrence analysis and Fisheries Application fisheries data Delay 4 Dimension 2

19 Recurrence analysis and Fisheries Application fisheries data Spatio-Temporal Entropy Delay 4 Dimension 2

20 Recurrence analysis and Fisheries Application fisheries data Delay 10 Dimension 6

21 Recurrence analysis and Fisheries Application fisheries data Spatio-Temporal Entropy Delay 10 Dimension 6