Causal Dynamic Time Lag (CDT) Applications to Space Weather Mandar Chandorkar 1 Multiscale Dynamics CWI, Amsterdam 2 TAU Research Unit INRIA, Paris-Saclay 25 September 2018 Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 1 / 47
47
Embed
Causal Dynamic Time Lag (CDT) - Inria › ... › 11 › MandarChandorkar-MDG-TAO-talk-sep… · Causal Dynamic Time Lag (CDT) Applications to Space Weather Mandar Chandorkar 1Multiscale
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Causal Dynamic Time Lag (CDT)Applications to Space Weather
Mandar Chandorkar
1Multiscale DynamicsCWI, Amsterdam
2TAU Research UnitINRIA, Paris-Saclay
25 September 2018
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 1 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 2 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 3 / 47
Figure: The Sun-Earth system
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 4 / 47
Figure: Effects of Solar Disturbances
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 5 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 6 / 47
Figure: Yield curves and Recessions in the U.S. Economy
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 7 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 8 / 47
Granger Causality
1 The cause happens prior to its effect.
2 The cause has unique information about the future values of its effect.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 9 / 47
Figure: By BiObserver - Own work, CC BY-SA 3.0,https://commons.wikimedia.org/w/index.php?curid=33470670
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 10 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 11 / 47
Nonparametric learning of time-lag
[Zhou and Sornette, 2006] formulated the problem as minimisation of themismatch between two time series.Inputs: X (t),Y (t), two time series.Learn: A mapping φ(t1) = t2 which minimises
ε(tt , t2) = |X (t1)− Y (t2)| (1)
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 12 / 47
Figure: An example of energy landscape EX ,Y given by (1) for two noisy timeseries and the corresponding optimal path wandering at the bottom of the valleysimilarly to a river. This optimal path defines the mapping t1→ t2 = φ(t1).
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 13 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 14 / 47
Causal Dynamic Time-lag Inference (CDT)
Given an input (cause) and output (effect) time series, predict
Magnitude of output signal. (What/How much)
When the effect will be observed in the output signal (When)
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 15 / 47
CDT: Formal Definition
Input Signal (Cause)
t ∈ R+
x(t) ∈ X
Output Signal (Effect)
f : X → Rg : X → R+
∆(t) = g [x(t)]
y(t + ∆(t)) = f [x(t)]
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 16 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 17 / 47
Model Specification
Input Patterns x(t)
Causal Time WindowLower Limit: ` ∈ N ∪ 0Upper Limit: `+ h : h ∈ N
Targets y(t + `), · · · , y(t + `+ h − 1)
Model OutputsPredictions y(t + `), · · · , y(t + `+ h − 1)Time Lag Probabilities p(t + `), · · · , p(t + `+ h − 1)
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 18 / 47
Training: Motivation
Balance two incentives
1 Generate accurate predictions for time windowy(t + `), · · · , y(t + `+ h − 1)
2 Learn time lag structure according to some intuition.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 19 / 47
Training Loss
L(y (1:M), y (1:M), p(1:M)) =λ1∑i ,m
1
2M(y
(m)i − y
(m)i )2(1 + p
(m)i )
+
λ2J (y (1:M), y (1:M), p(1:M))
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 20 / 47
Training Loss Contd.
The term J (y (1:M), y (1:M), p(1:M)) penalizes the predicted probabilitiesp(1:M), for deviation from some chosen target probability.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 21 / 47
Target Probability
The target probability p for a time window [t + `, t + `+ h − 1] can becharacterized by:
Conjecture: Causal Time Lag
The lagged output y(t + i) which has greater predictability given x(t), is amore likely causal link.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 22 / 47
Target Probability Contd.
1 The target probability distribution for the time lag is,p(m) = softmax((y (m) − y (m))2/T )
2 The term J (y (1:M), y (1:M), p(1:M)) can be computed as the Hellingerdistance between p and p.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 23 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 24 / 47
Need for Benchmarks
1 Labelled data sets are small in size.
2 Most real world data sets dont have explicit labels for causal time lag.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 25 / 47
Benchmark Problems
x(t + 1) = (1− β)x(t) +N (0, σ2)
y(t + ∆(t)) = α||x(t)||2
Problem I: Constant Lag
∆(t) = k
Problem II: Constant Velocity ||x(t)||2; Fixed Distance d
∆(t) = d/(α||x(t)||2)
Problem III: Constant Acceleration a; Fixed Distance d
∆(t) = (√α2||x(t)||4 − 2ad − α||x(t)||2)/a
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 26 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 27 / 47
Data
Figure: Generated Data
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 28 / 47
Errors
Figure: Error in Output vs Error in Time Lag
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 29 / 47
Predictions
Figure: Test Set Time Series vs Predictions
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 30 / 47
Performance
Output
MAE: 8.602Pearson Corr: 0.964Spearman Corr: 0.999
Time Lag
MAE: 0Pearson Corr: N.ASpearman Corr: 1.0
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 31 / 47
Outline
1 MotivationSpace WeatherFinancial Markets
2 Causality in Time SeriesConceptsExisting Research
3 Problem & ModelDescriptionProposed Solution
4 ApplicationsBenchmarksProblem IProblem IIProblem III
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 32 / 47
Data
Figure: Generated Data
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 33 / 47
Time Lags
Figure: Time Lags
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 34 / 47
Test Data Distribution
Figure: Output vs Time Lags
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 35 / 47
Predictions
Figure: Test Set Time Series vs Predictions
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 36 / 47
Errors
Figure: Error in Output vs Error in Time Lag
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 37 / 47
Zhou, W.-X. and Sornette, D. (2006).Non-parametric determination of real-time lag structure between twotime series: The optimal thermal causal path method withapplications to economic data.Journal of Macroeconomics, 28(1):195 – 224.Nonlinear Macroeconomic Dynamics.
Mandar Chandorkar (CWI & INRIA) Causal Dynamic Time Lag (CDT) 25 September 2018 47 / 47