Soft-DTW: a Differentiable Loss Function for Time-Series. Appendix material

Size: px
Start display at page:

Download "Soft-DTW: a Differentiable Loss Function for Time-Series. Appendix material"

Transcription

1 : a Differentiable Loss Function for Time-Series Appendix material A. Recursive forward computation of the average path matrix The average alignment under Gibbs distributionp γ can be computed with the following forward recurrence, which mimics closely Bellman s original recursion. For eachi n,j m, define E i+,j+ = [ ] e δ i+,j+/γ E i,j i T j e ri+,j+/γ + [ ] [ e δ i+,j+/γ E i,j+ T j e ri+,j+/γ + e δi+,j+/γ E i+,j ] i e rij/γ Here terms r ij are computed following the recursion in Algorithm. Border matrices are initialized to, except for E, which is initialized to[]. Upon completion, the average alignment matrix is stored ine n,m. The operation above consists in summing three matrices of size (i +,j + ). There are exactly (nm) such updates. A careful implementation of this algorithm, that would only store two arrays of matrices, as Algorithm only store two arrays of values, can be carried out innmmin(n,m) space but it would still require(nm) operations.

2 : a Differentiable Loss Function for Time-Series B. Barycenters obtained with random initialization ( =) ( =.) (a) CBF ( =) ( =.) (b) Herring ( =) ( =.) (c) Medical Images ( =) ( =.) (d) Synthetic Control ( =) ( =.) (e) Wave Gesture Library Y

3 : a Differentiable Loss Function for Time-Series C. Barycenters obtained with mean initialization ( =) ( =.) (a) CBF ( =) ( =.) (b) Herring ( =) ( =.) (c) Medical Images ( =) ( =.) (d) Synthetic Control ( =) ( =.) (e) Wave Gesture Library Y

4 D. More interpolation results : a Differentiable Loss Function for Time-Series Left: results obtained under loss. Right: results obtained under soft-dtw (γ = ) loss (a) ArrowHead (b) ECG 5 5 (c) ItalyPowerDemand (d) TwoLeadECG

5 : a Differentiable Loss Function for Time-Series E. Clusters obtained byk-means under DTW or soft-dtw geometry CBF dataset Cluster (8 points) Cluster (9 points) Cluster ( points) (a) (γ =, random initialization) Cluster (8 points) Cluster (8 points) (b) (γ =, mean initialization) Cluster ( points) Cluster (8 points) Cluster ( points) Cluster ( points) (c) (random initialization) Cluster ( points) Cluster ( points) (d) ( mean initialization) Cluster ( points)

6 : a Differentiable Loss Function for Time-Series ECG dataset Cluster (59 points) Cluster ( points) (a) (γ =, random initialization) Cluster (8 points) Cluster (9 points) (b) (γ =, mean initialization) Cluster (8 points) Cluster (7 points) 6 8 (c) (random initialization) 6 8 Cluster (76 points) Cluster ( points) 6 8 (d) ( mean initialization) 6 8

7 F. More visualizations of time-series prediction : a Differentiable Loss Function for Time-Series (a) CBF (b) ECG (c) ECG (d) ShapesAll (e) uwavegesturelibrary Y

8 : a Differentiable Loss Function for Time-Series G. Barycenters: DTW loss (Eq. withγ = ) achieved with random init Dataset γ = γ =. γ =. γ =. Subgradient method mean 5words Adiac ArrowHead Beef BeetleFly BirdChicken CBF Car ChlorineConcentration CinC ECG torso Coffee Computers Cricket X Cricket Y Cricket Z DiatomSizeReduction DistalPhalanxOutlineAgeGroup DistalPhalanxOutlineCorrect DistalPhalanxTW ECG ECG ECGFiveDays Earthquakes ElectricDevices FISH FaceAll FaceFour FacesUCR FordA FordB Gun Point Ham HandOutlines Haptics Herring InlineSkate N/A N/A N/A N/A ItalyPowerDemand MedicalImages MiddlePhalanxOutlineAgeGroup MiddlePhalanxOutlineCorrect MiddlePhalanxTW MoteStrain NonInvasiveFatalECG Thorax NonInvasiveFatalECG Thorax OSULeaf OliveOil PhalangesOutlinesCorrect Phoneme Plane ProximalPhalanxOutlineAgeGroup ProximalPhalanxOutlineCorrect ProximalPhalanxTW RefrigerationDevices ScreenType ShapeletSim ShapesAll SmallKitchenAppliances SonyAIBORobotSurface SonyAIBORobotSurfaceII StarLightCurves Strawberry SwedishLeaf Symbols ToeSegmentation ToeSegmentation Trace TwoLeadECG Two Patterns UWaveGestureLibraryAll Wine WordsSynonyms Worms WormsTwoClass synthetic control uwavegesturelibrary X uwavegesturelibrary Y uwavegesturelibrary Z wafer yoga

9 : a Differentiable Loss Function for Time-Series H. Barycenters: DTW loss (Eq. () withγ = ) achieved with init Dataset γ = γ =. γ =. γ =. Subgradient method mean 5words Adiac ArrowHead Beef BeetleFly BirdChicken CBF Car ChlorineConcentration CinC ECG torso Coffee Computers Cricket X Cricket Y Cricket Z DiatomSizeReduction DistalPhalanxOutlineAgeGroup DistalPhalanxOutlineCorrect DistalPhalanxTW ECG ECG ECGFiveDays Earthquakes ElectricDevices FISH FaceAll FaceFour FacesUCR FordA FordB Gun Point Ham HandOutlines Haptics Herring InlineSkate N/A N/A N/A ItalyPowerDemand MedicalImages MiddlePhalanxOutlineAgeGroup MiddlePhalanxOutlineCorrect MiddlePhalanxTW MoteStrain NonInvasiveFatalECG Thorax NonInvasiveFatalECG Thorax OSULeaf OliveOil PhalangesOutlinesCorrect Phoneme Plane ProximalPhalanxOutlineAgeGroup ProximalPhalanxOutlineCorrect ProximalPhalanxTW RefrigerationDevices ScreenType ShapeletSim ShapesAll SmallKitchenAppliances SonyAIBORobotSurface SonyAIBORobotSurfaceII StarLightCurves Strawberry SwedishLeaf Symbols ToeSegmentation ToeSegmentation Trace TwoLeadECG Two Patterns UWaveGestureLibraryAll Wine WordsSynonyms Worms WormsTwoClass synthetic control uwavegesturelibrary X uwavegesturelibrary Y uwavegesturelibrary Z wafer yoga

10 : a Differentiable Loss Function for Time-Series I.k-means clustering: DTW loss achieved (Eq. (5) withγ =, log-scaled) when using random initialization Dataset γ = γ =. γ =. γ =. Subgradient method mean 5words Adiac ArrowHead Beef BeetleFly BirdChicken CBF Car ChlorineConcentration CinC ECG torso Coffee Computers Cricket X Cricket Y Cricket Z DiatomSizeReduction DistalPhalanxOutlineAgeGroup DistalPhalanxOutlineCorrect DistalPhalanxTW ECG ECG ECGFiveDays Earthquakes ElectricDevices FISH FaceAll FaceFour FacesUCR FordA FordB Gun Point Ham MedicalImages MiddlePhalanxOutlineAgeGroup MiddlePhalanxOutlineCorrect MiddlePhalanxTW MoteStrain NonInvasiveFatalECG Thorax 7.78 N/A N/A N/A N/A N/A N/A ProximalPhalanxTW RefrigerationDevices ScreenType ShapeletSim ShapesAll SmallKitchenAppliances SonyAIBORobotSurface SonyAIBORobotSurfaceII StarLightCurves N/A N/A N/A N/A Trace TwoLeadECG Two Patterns UWaveGestureLibraryAll Wine WordsSynonyms Worms WormsTwoClass synthetic control uwavegesturelibrary X N/A N/A N/A N/A N/A

11 : a Differentiable Loss Function for Time-Series J.k-means clustering: DTW loss achieved (Eq. (5) withγ =, log-scaled) when using mean initialization Dataset γ = γ =. γ =. γ =. Subgradient method mean 5words Adiac ArrowHead Beef BeetleFly BirdChicken CBF Car ChlorineConcentration CinC ECG torso Coffee Computers Cricket X Cricket Y Cricket Z DiatomSizeReduction DistalPhalanxOutlineAgeGroup DistalPhalanxOutlineCorrect DistalPhalanxTW ECG ECG ECGFiveDays Earthquakes ElectricDevices FISH FaceAll FaceFour FacesUCR FordA FordB Gun Point Ham HandOutlines.7 N/A N/A N/A N/A N/A N/A MedicalImages MiddlePhalanxOutlineAgeGroup MiddlePhalanxOutlineCorrect MiddlePhalanxTW MoteStrain NonInvasiveFatalECG Thorax N/A N/A N/A N/A N/A N/A ProximalPhalanxTW RefrigerationDevices ScreenType ShapeletSim ShapesAll SmallKitchenAppliances SonyAIBORobotSurface SonyAIBORobotSurfaceII StarLightCurves N/A N/A N/A N/A Trace TwoLeadECG Two Patterns UWaveGestureLibraryAll Wine WordsSynonyms Worms WormsTwoClass synthetic control uwavegesturelibrary X N/A N/A N/A N/A N/A

12 : a Differentiable Loss Function for Time-Series K. Time-series prediction: DTW loss achieved when using random init Dataset loss γ = γ =. γ =. γ =. loss 5words Adiac ArrowHead Beef BeetleFly BirdChicken CBF Car ChlorineConcentration CinC ECG torso Coffee Computers Cricket X Cricket Y Cricket Z DiatomSizeReduction DistalPhalanxOutlineAgeGroup DistalPhalanxOutlineCorrect DistalPhalanxTW ECG ECG ECGFiveDays Earthquakes ElectricDevices FISH FaceAll FaceFour FacesUCR Gun Point Ham Haptics Herring InsectWingbeatSound ItalyPowerDemand LargeKitchenAppliances Lighting Lighting Meat MedicalImages MiddlePhalanxOutlineAgeGroup MiddlePhalanxOutlineCorrect MiddlePhalanxTW MoteStrain NonInvasiveFatalECG Thorax OSULeaf OliveOil PhalangesOutlinesCorrect Phoneme Plane ProximalPhalanxOutlineAgeGroup ProximalPhalanxOutlineCorrect ProximalPhalanxTW RefrigerationDevices ShapeletSim ShapesAll SonyAIBORobotSurface SonyAIBORobotSurfaceII Strawberry SwedishLeaf Symbols ToeSegmentation ToeSegmentation Trace TwoLeadECG Two Patterns UWaveGestureLibraryAll Wine WordsSynonyms Worms WormsTwoClass synthetic control uwavegesturelibrary X uwavegesturelibrary Y uwavegesturelibrary Z wafer yoga

13 : a Differentiable Loss Function for Time-Series L. Time-series prediction: DTW loss achieved when using init Dataset loss γ = γ =. γ =. γ =. loss 5words Adiac ArrowHead Beef BeetleFly BirdChicken CBF Car ChlorineConcentration CinC ECG torso Coffee Computers Cricket X Cricket Y Cricket Z DiatomSizeReduction DistalPhalanxOutlineAgeGroup DistalPhalanxOutlineCorrect DistalPhalanxTW ECG ECG ECGFiveDays Earthquakes ElectricDevices FISH FaceAll FaceFour FacesUCR Gun Point Ham Haptics Herring InsectWingbeatSound ItalyPowerDemand LargeKitchenAppliances Lighting Lighting Meat MedicalImages MiddlePhalanxOutlineAgeGroup MiddlePhalanxOutlineCorrect MiddlePhalanxTW MoteStrain NonInvasiveFatalECG Thorax OSULeaf OliveOil PhalangesOutlinesCorrect Phoneme Plane ProximalPhalanxOutlineAgeGroup ProximalPhalanxOutlineCorrect ProximalPhalanxTW RefrigerationDevices ShapeletSim ShapesAll SonyAIBORobotSurface SonyAIBORobotSurfaceII Strawberry SwedishLeaf Symbols ToeSegmentation ToeSegmentation Trace TwoLeadECG Two Patterns UWaveGestureLibraryAll Wine WordsSynonyms Worms WormsTwoClass synthetic control uwavegesturelibrary X uwavegesturelibrary Y uwavegesturelibrary Z wafer yoga

arxiv: v2 [stat.ml] 20 Feb 2018

arxiv: v2 [stat.ml] 20 Feb 2018 Marco Cuturi Mathieu Blondel arxiv:7.5v [stat.ml] Feb 8 Abstract We propose in this paper a differentiable learning loss between time series, building upon the celebrated dynamic time warping (DTW) discrepancy.

More information

Time Series Classification in Dissimilarity Spaces

Time Series Classification in Dissimilarity Spaces Proceedings 1st International Workshop on Advanced Analytics and Learning on Temporal Data AALTD 2015 Time Series Classification in Dissimilarity Spaces Brijnesh J. Jain and Stephan Spiegel Berlin Institute

More information

arxiv: v1 [cs.ds] 26 Jan 2017

arxiv: v1 [cs.ds] 26 Jan 2017 Fast and Accurate Series Classifition with Patrick Schäfer Humboldt University of Berlin Berlin, Germany patrick.schaefer@hu-berlin.de Ulf Leser Humboldt University of Berlin Berlin, Germany leser@informatik.hu-berlin.de

More information

Extracting Statistical Graph Features for Accurate and Efficient Time Series Classification

Extracting Statistical Graph Features for Accurate and Efficient Time Series Classification Extracting Statistical Graph Features for Accurate and Efficient Time Series Classification Daoyuan Li University of Luxembourg Luxembourg City, Luxembourg daoyuan.li@uni.lu Jessica Lin George Mason University

More information

LSTM Fully Convolutional Networks for Time Series Classification

LSTM Fully Convolutional Networks for Time Series Classification 1 LSTM Fully Convolutional Networks for Time Series Classification Fazle Karim 1, Somshubra Majumdar 2, Houshang Darabi 1, Senior Member, IEEE, and Shun Chen 1 arxiv:1709.05206v1 [cs.lg] 8 Sep 2017 Abstract

More information

arxiv: v1 [cs.lg] 23 Jun 2017

arxiv: v1 [cs.lg] 23 Jun 2017 TimeNet: Pre-trained deep recurrent neural network for time series classification arxiv:1706.08838v1 [cs.lg] 23 Jun 2017 Pankaj Malhotra Vishnu TV Lovekesh Vig Puneet Agarwal Gautam Shroff TCS Research,

More information

RPM: Representative Pattern Mining for Efficient Time Series Classification

RPM: Representative Pattern Mining for Efficient Time Series Classification RPM: Representative Pattern Mining for Efficient Time Series Classification Xing Wang, Jessica Lin George Mason University Dept. of Computer Science xwang4@gmu.edu, jessica@gmu.edu Pavel Senin University

More information

Fast and Accurate Time Series Classification with WEASEL

Fast and Accurate Time Series Classification with WEASEL Fast and Accurate Series Classifition with ABRACT Patrick Schäfer Humboldt University of Berlin, Germany patrick.schaefer@hu-berlin.de series (TS) occur in many scientific and commercial applitions, ranging

More information

On Time-series Topological Data Analysis: New Data and Opportunities

On Time-series Topological Data Analysis: New Data and Opportunities On Time-series Topological Data Analysis: New Data and Opportunities Lee M. Seversky Air Force Research Laboratory Lee.Seversky@us.af.mil Shelby Davis Black River Systems Davis@brsc.com Matthew Berger

More information

arxiv: v1 [cs.lg] 16 Jan 2014

arxiv: v1 [cs.lg] 16 Jan 2014 An Empirical Evaluation of Similarity Measures for Time Series Classification arxiv:1401.3973v1 [cs.lg] 16 Jan 2014 Abstract Joan Serrà, Josep Ll. Arcos Artificial Intelligence Research Institute (IIIA-CSIC),

More information

INSIGHT: Efficient and Effective Instance Selection for Time-Series Classification

INSIGHT: Efficient and Effective Instance Selection for Time-Series Classification INSIGHT: Efficient and Effective Instance Selection for Time-Series Classification Krisztian Buza, Alexandros Nanopoulos, and Lars Schmidt-Thieme Information Systems and Machine Learning Lab (ISMLL), University

More information

arxiv: v1 [cs.lg] 18 Sep 2018

arxiv: v1 [cs.lg] 18 Sep 2018 From BOP to BOSS and Beyond: Time Series Classification with Dictionary Based Classifiers arxiv:1809.06751v1 [cs.lg] 18 Sep 2018 James Large 1, Anthony Bagnall 1 Simon Malinowski 2 Romain Tavenard 3 1

More information

Hubness-aware Classification, Instance Selection and Feature Construction: Survey and Extensions to Time-Series

Hubness-aware Classification, Instance Selection and Feature Construction: Survey and Extensions to Time-Series Hubness-aware Classification, Instance Selection and Feature Construction: Survey and Extensions to Time-Series Nenad Tomašev, Krisztian Buza, Kristóf Marussy, and Piroska B. Kis Abstract Time-series classification

More information

Extracting texture features for time series classification

Extracting texture features for time series classification Universidade de São Paulo Biblioteca Digital da Produção Intelectual - BDPI Departamento de Ciências de Computação - ICMC/SCC Comunicações em Eventos - ICMC/SCC 2014-08 Extracting texture features for

More information

Applying Nature-Inspired Optimization Algorithms for Selecting Important Timestamps to Reduce Time Series Dimensionality

Applying Nature-Inspired Optimization Algorithms for Selecting Important Timestamps to Reduce Time Series Dimensionality Applying Nature-Inspired Optimization Algorithms for Selecting Important Timestamps to Reduce Time Series Dimensionality Muhammad Marwan Muhammad Fuad Coventry University School of Computing, Electronics

More information

Comparison of different weighting schemes for the knn classifier on time-series data

Comparison of different weighting schemes for the knn classifier on time-series data Comparison of different weighting schemes for the knn classifier on time-series data Zoltan Geler 1, Vladimir Kurbalija 2, Miloš Radovanović 2, Mirjana Ivanović 2 1 Faculty of Philosophy, University of

More information

Time-Series Classification based on Individualised Error Prediction

Time-Series Classification based on Individualised Error Prediction Time-Series Classification based on Individualised Error Prediction Krisztian Buza, Alexandros Nanopoulos, Lars Schmidt-Thieme Information Systems and Machine Learning Lab University of Hildesheim Hildesheim,

More information

Invariant Time-Series Classification

Invariant Time-Series Classification Invariant Time-Series Classification Josif Grabocka 1, Alexandros Nanopoulos 2, and Lars Schmidt-Thieme 1 1 Information Systems and Machine Learning Lab Samelsonplatz 22, 31141 Hildesheim, Germany 2 University

More information

IQ Estimation for Accurate Time-Series Classification

IQ Estimation for Accurate Time-Series Classification IQ Estimation for Accurate Time-Series Classification Krisztian Buza, Alexandros Nanopoulos, Lars Schmidt-Thieme Information Systems and Machine Learning Lab University of Hildesheim Hildesheim, Germany

More information

arxiv: v1 [stat.ml] 19 Aug 2015

arxiv: v1 [stat.ml] 19 Aug 2015 Time Series Clustering via Community Detection in Networks Leonardo N. Ferreira a,, Liang Zhao b a Institute of Mathematics and Computer Science, University of São Paulo Av. Trabalhador São-carlense, 400

More information

Model-based Time Series Classification

Model-based Time Series Classification Model-based Time Series Classification Alexios Kotsifakos 1 and Panagiotis Papapetrou 2 1 Department of Computer Science and Engineering, University of Texas at Arlington, USA 2 Department of Computer

More information

A Shapelet Learning Method for Time Series Classification

A Shapelet Learning Method for Time Series Classification 2016 IEEE 28th International Conference on Tools with Artificial Intelligence A Shapelet Learning Method for Time Series Classification Yi Yang 1, Qilin Deng 1, Furao Shen 1,, Jinxi Zhao 1 and Chaomin

More information

Jon Hills, Jason Lines, Edgaras Baranauskas, James Mapp & Anthony Bagnall

Jon Hills, Jason Lines, Edgaras Baranauskas, James Mapp & Anthony Bagnall Classification of time series by shapelet transformation Jon Hills, Jason Lines, Edgaras Baranauskas, James Mapp & Anthony Bagnall Data Mining and Knowledge Discovery ISSN 1384-5810 Data Min Knowl Disc

More information

Adaptive Feature Based Dynamic Time Warping

Adaptive Feature Based Dynamic Time Warping 264 IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.1, January 2010 Adaptive Feature Based Dynamic Time Warping Ying Xie and Bryan Wiltgen Department of Computer Science

More information

Pattern Recognition. A global averaging method for dynamic time warping, with applications to clustering

Pattern Recognition. A global averaging method for dynamic time warping, with applications to clustering Pattern Recognition 44 (2) 678 693 Contents lists available at ScienceDirect Pattern Recognition journal homepage: www.elsevier.com/locate/pr A global averaging method for dynamic time warping, with applications

More information

Clustering of large time series datasets

Clustering of large time series datasets Intelligent Data Analysis 18 (2014) 793 817 793 DOI 10.3233/IDA-140669 IOS Press Clustering of large time series datasets Saeed Aghabozorgi and Teh Ying Wah Faculty of Computer Science and Information

More information

Multi-Scale Convolutional Neural Networks for Time Series Classification

Multi-Scale Convolutional Neural Networks for Time Series Classification Multi-Scale Convolutional Neural Networks for Time Series Classification Zhicheng Cui Department of Computer Science and Engineering Washington University in St. Louis, USA z.cui@wustl.edu Wenlin Chen

More information

EXPLORING CLASSIFICATION BY DISCRIMINATIVE INTERPOLATION

EXPLORING CLASSIFICATION BY DISCRIMINATIVE INTERPOLATION EXPLORING CLASSIFICATION BY DISCRIMINATIVE INTERPOLATION By Kathryn Sarullo A SENIOR RESEARCH PAPER PRESENTED TO THE DEPARTMENT OF MATHEMATICS AND COMPUTER SCIENCE OF STETSON UNIVERSITY IN PARTIAL FULFILLMENT

More information

Efficient Dynamic Time Warping for Time Series Classification

Efficient Dynamic Time Warping for Time Series Classification Indian Journal of Science and Technology, Vol 9(, DOI: 0.7485/ijst/06/v9i/93886, June 06 ISSN (Print : 0974-6846 ISSN (Online : 0974-5645 Efficient Dynamic Time Warping for Time Series Classification Kumar

More information

Time Series Classification by Sequence Learning in All-Subsequence Space

Time Series Classification by Sequence Learning in All-Subsequence Space Time Series Classification by Sequence Learning in All-Subsequence Space Thach Le Nguyen Insight Centre for Data Analytics University College Dublin, Ireland thach.lenguyen@insight-centre.org Severin Gsponer

More information

Dynamic Time Warping Under Limited Warping Path Length

Dynamic Time Warping Under Limited Warping Path Length Dynamic Time Warping Under Limited Warping Path Length Zheng Zhang, Romain Tavenard, Adeline Bailly, Xiaotong Tang, Ping Tang, Thomas Corpetti To cite this version: Zheng Zhang, Romain Tavenard, Adeline

More information

Learning DTW-Shapelets for Time-Series Classification

Learning DTW-Shapelets for Time-Series Classification Learning DTW-Shapelets for -Series Classification Mit Shah 1, Josif Grabocka 1, Nicolas Schilling 1, Martin Wistuba 1, Lars Schmidt-Thieme 1 1 Information Systems and Machine Learning Lab University of

More information

Exact and Approximate Reverse Nearest Neighbor Search for Multimedia Data

Exact and Approximate Reverse Nearest Neighbor Search for Multimedia Data Exact and Approximate Reverse Nearest Neighbor Search for Multimedia Data Jessica Lin David Etter David DeBarr jessica@ise.gmu.edu detter@gmu.edu Dave.DeBarr@microsoft.com George Mason University Microsoft

More information

Soft-DTW: a Differentiable Loss Function for Time-Series

Soft-DTW: a Differentiable Loss Function for Time-Series Marco Cuturi 1 Mathieu Blondel 2 Abstract We propose in this paper a differentiable learning loss between time series, building upon the celebrated dynamic time warping (DTW) discrepancy. Unlike the Euclidean

More information

Time Series Clustering Ensemble Algorithm Based on Locality Preserving Projection

Time Series Clustering Ensemble Algorithm Based on Locality Preserving Projection Based on Locality Preserving Projection 2 Information & Technology College, Hebei University of Economics & Business, 05006 Shijiazhuang, China E-mail: 92475577@qq.com Xiaoqing Weng Information & Technology

More information

Similarity Matching for Uncertain Time Series: Analytical and Experimental Comparison

Similarity Matching for Uncertain Time Series: Analytical and Experimental Comparison Similarity Matching for Uncertain Time Series: Analytical and Experimental Comparison Michele Dallachiesa University of Trento, Italy dallachiesa@disi.unitn.it Besmira Nushi University of Trento, Italy

More information

Time Warping Symbolic Aggregation Approximation with Bag-of-Patterns Representation for Time Series Classification

Time Warping Symbolic Aggregation Approximation with Bag-of-Patterns Representation for Time Series Classification Time Warping Symbolic Aggregation Approximation with Bag-of-Patterns Representation for Time Series Classification Zhiguang Wang Department of Computer Science and Electric Engineering University of Maryland

More information

arxiv: v1 [cs.lg] 4 Feb 2016

arxiv: v1 [cs.lg] 4 Feb 2016 The Great Time Series Classification Bake Off: An Experimental Evaluation of Recently Proposed Algorithms. Extended Version arxiv:1602.01711v1 [cs.lg] 4 Feb 2016 Anthony Bagnall University of East Anglia

More information

Encoding Time Series as Images for Visual Inspection and Classification Using Tiled Convolutional Neural Networks

Encoding Time Series as Images for Visual Inspection and Classification Using Tiled Convolutional Neural Networks Trajectory-Based Behavior Analytics: Papers from the 2015 AAAI Workshop Encoding Time Series as Images for Visual Inspection and Classification Using Tiled Convolutional Neural Networks Zhiguang Wang and

More information

Bag-of-Temporal-SIFT-Words for Time Series Classification

Bag-of-Temporal-SIFT-Words for Time Series Classification Proceedings st International Workshop on Advanced Analytics and Learning on Temporal Data AALTD 5 Bag-of-Temporal-SIFT-Words for Time Series Classification Adeline Bailly, Simon Malinowski, Romain Tavenard,

More information

arxiv: v1 [cs.lg] 24 Sep 2015

arxiv: v1 [cs.lg] 24 Sep 2015 Spatially Encoding Temporal Correlations to Classify Temporal Data Using Convolutional Neural Networks Zhiguang Wang a,, Tim Oates a arxiv:1509.07481v1 [cs.lg] 24 Sep 2015 a Department of Computer Science

More information

arxiv: v1 [cs.lg] 1 Jun 2015

arxiv: v1 [cs.lg] 1 Jun 2015 Imaging Time-Series to Improve Classification and Imputation Zhiguang Wang and Tim Oates Department of Computer Science and Electric Engineering University of Maryland, Baltimore County {stephen.wang,

More information

Efficient Pattern-Based Time Series Classification on GPU

Efficient Pattern-Based Time Series Classification on GPU Efficient Pattern-Based Time Series Classification on GPU Kai-Wei Chang 1, Biplab Deka 2, Wen-Mei W. Hwu 2, Dan Roth 1 1 Dept. of Computer Science, 2 Dept. of Electrical and Computer Engineering University

More information

INTRODUCTION TO CLUSTER ALGEBRAS

INTRODUCTION TO CLUSTER ALGEBRAS INTRODUCTION TO CLUSTER ALGEBRAS NAN LI (MIT) Cluster algebras are a class of commutative ring, introduced in 000 by Fomin and Zelevinsky, originally to study Lusztig s dual canonical basis and total positivity.

More information

D-BAUG Informatik I. Exercise session: week 5 HS 2018

D-BAUG Informatik I. Exercise session: week 5 HS 2018 1 D-BAUG Informatik I Exercise session: week 5 HS 2018 Homework 2 Questions? Matrix and Vector in Java 3 Vector v of length n: Matrix and Vector in Java 3 Vector v of length n: double[] v = new double[n];

More information

Quaternions & Rotation in 3D Space

Quaternions & Rotation in 3D Space Quaternions & Rotation in 3D Space 1 Overview Quaternions: definition Quaternion properties Quaternions and rotation matrices Quaternion-rotation matrices relationship Spherical linear interpolation Concluding

More information

Longest Common Subsequences and Substrings

Longest Common Subsequences and Substrings Longest Common Subsequences and Substrings Version November 5, 2014 Version November 5, 2014 Longest Common Subsequences and Substrings 1 / 16 Longest Common Subsequence Given two sequences X = (x 1, x

More information

arxiv: v1 [stat.ml] 17 Sep 2016

arxiv: v1 [stat.ml] 17 Sep 2016 ADAGIO: Fast Data-aware Near-Isometric Linear Embeddings Jaros law B lasiok Charalampos E. Tsourakakis September 2, 216 arxiv:169.5388v1 [stat.ml] 17 Sep 216 Abstract Many important applications, including

More information

Enhancing Time Series Clustering by Incorporating Multiple Distance Measures with Semi-Supervised Learning

Enhancing Time Series Clustering by Incorporating Multiple Distance Measures with Semi-Supervised Learning Zhou J, Zhu SF, Huang X et al. Enhancing time series clustering by incorporating multiple distance measures with semisupervised learning. JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY 30(4): 859 873 July

More information

Improving SVM Classification on Imbalanced Data Sets in Distance Spaces

Improving SVM Classification on Imbalanced Data Sets in Distance Spaces Improving SVM Classification on Imbalanced Data Sets in Distance Spaces Suzan Köknar-Tezel Department of Computer Science Saint Joseph s University Philadelphia, USA tezel@sju.edu Longin Jan Latecki Dept.

More information

Non-exhaustive, Overlapping k-means

Non-exhaustive, Overlapping k-means Non-exhaustive, Overlapping k-means J. J. Whang, I. S. Dhilon, and D. F. Gleich Teresa Lebair University of Maryland, Baltimore County October 29th, 2015 Teresa Lebair UMBC 1/38 Outline Introduction NEO-K-Means

More information

SpADe: On Shape-based Pattern Detection in Streaming Time Series

SpADe: On Shape-based Pattern Detection in Streaming Time Series : On Shape-based Pattern etection in Streaming Time Series Yueguo Chen Mario A Nascimento Beng Chin Ooi Anthony K H Tung National University of Singapore {chenyueg, ooibc, atung}@compnusedusg University

More information

Parallelizing The Matrix Multiplication. 6/10/2013 LONI Parallel Programming Workshop

Parallelizing The Matrix Multiplication. 6/10/2013 LONI Parallel Programming Workshop Parallelizing The Matrix Multiplication 6/10/2013 LONI Parallel Programming Workshop 2013 1 Serial version 6/10/2013 LONI Parallel Programming Workshop 2013 2 X = A md x B dn = C mn d c i,j = a i,k b k,j

More information

Efficient Proper Length Time Series Motif Discovery

Efficient Proper Length Time Series Motif Discovery 3 IEEE 3th International Conference on Data Mining Efficient Proper Length Time Series Motif Discovery Sorrachai Yingchareonthaornchai, Haemaan Sivaraks, Thanain Rakthanmanon*, Chotirat Ann Ratanamahatana

More information

SAX-VSM: Interpretable Time Series Classification Using SAX and Vector Space Model

SAX-VSM: Interpretable Time Series Classification Using SAX and Vector Space Model SAX-VSM: Interpretable Time Series Classification Using SAX and Vector Space Model Pavel Senin Information and Computer Sciences Department, University of Hawaii at Manoa, Honolulu, HI, 96822 senin@hawaii.edu

More information

Efficient Subsequence Search on Streaming Data Based on Time Warping Distance

Efficient Subsequence Search on Streaming Data Based on Time Warping Distance 2 ECTI TRANSACTIONS ON COMPUTER AND INFORMATION TECHNOLOGY VOL.5, NO.1 May 2011 Efficient Subsequence Search on Streaming Data Based on Time Warping Distance Sura Rodpongpun 1, Vit Niennattrakul 2, and

More information

Name Hr. Honors Geometry Lesson 9-1: Translate Figures and Use Vectors

Name Hr. Honors Geometry Lesson 9-1: Translate Figures and Use Vectors Name Hr Honors Geometry Lesson 9-1: Translate Figures and Use Vectors Learning Target: By the end of today s lesson we will be able to successfully use a vector to translate a figure. Isometry: An isometry

More information

Bag-of-Temporal-SIFT-Words for Time Series Classification

Bag-of-Temporal-SIFT-Words for Time Series Classification Bag-of-Temporal-SIFT-Words for Time Series Classification Adeline Bailly, Simon Malinowski, Romain Tavenard, Thomas Guyet, Laetitia Chapel To cite this version: Adeline Bailly, Simon Malinowski, Romain

More information

Opening the Black Box Data Driven Visualizaion of Neural N

Opening the Black Box Data Driven Visualizaion of Neural N Opening the Black Box Data Driven Visualizaion of Neural Networks September 20, 2006 Aritificial Neural Networks Limitations of ANNs Use of Visualization (ANNs) mimic the processes found in biological

More information

12 Edit distance. Summer Term 2011

12 Edit distance. Summer Term 2011 12 Edit distance Summer Term 2011 Jan-Georg Smaus Problem: similarity of strings Edit distance For two strings A and B, compute, as efficiently as possible, the edit distance D(A,B) and a minimal sequence

More information

Algorithm Design and Analysis

Algorithm Design and Analysis Algorithm Design and Analysis LECTURE 13 Divide and Conquer Closest Pair of Points Convex Hull Strassen Matrix Mult. Adam Smith 9/24/2008 A. Smith; based on slides by E. Demaine, C. Leiserson, S. Raskhodnikova,

More information

Machine Learning (BSMC-GA 4439) Wenke Liu

Machine Learning (BSMC-GA 4439) Wenke Liu Machine Learning (BSMC-GA 4439) Wenke Liu 01-31-017 Outline Background Defining proximity Clustering methods Determining number of clusters Comparing two solutions Cluster analysis as unsupervised Learning

More information

The Pixel Array method for solving nonlinear systems

The Pixel Array method for solving nonlinear systems The Pixel Array method for solving nonlinear systems David I. Spivak Joint with Magdalen R.C. Dobson, Sapna Kumari, and Lawrence Wu dspivak@math.mit.edu Mathematics Department Massachusetts Institute of

More information

A Practical Review of Uniform B-Splines

A Practical Review of Uniform B-Splines A Practical Review of Uniform B-Splines Kristin Branson A B-spline is a convenient form for representing complicated, smooth curves. A uniform B-spline of order k is a piecewise order k Bezier curve, and

More information

CS354 Computer Graphics Ray Tracing. Qixing Huang Januray 24th 2017

CS354 Computer Graphics Ray Tracing. Qixing Huang Januray 24th 2017 CS354 Computer Graphics Ray Tracing Qixing Huang Januray 24th 2017 Graphics Pipeline Elements of rendering Object Light Material Camera Geometric optics Modern theories of light treat it as both a wave

More information

Jane Li. Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute

Jane Li. Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute Jane Li Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute (3 pts) Compare the testing methods for testing path segment and finding first

More information

Active Clustering and Ranking

Active Clustering and Ranking Active Clustering and Ranking Rob Nowak, University of Wisconsin-Madison IMA Workshop on "High-Dimensional Phenomena" (9/26-30, 2011) Gautam Dasarathy Brian Eriksson (Madison/Boston) Kevin Jamieson Aarti

More information

Greedy Algorithms. Algorithms

Greedy Algorithms. Algorithms Greedy Algorithms Algorithms Greedy Algorithms Many algorithms run from stage to stage At each stage, they make a decision based on the information available A Greedy algorithm makes decisions At each

More information

Rasterization. COMP 575/770 Spring 2013

Rasterization. COMP 575/770 Spring 2013 Rasterization COMP 575/770 Spring 2013 The Rasterization Pipeline you are here APPLICATION COMMAND STREAM 3D transformations; shading VERTEX PROCESSING TRANSFORMED GEOMETRY conversion of primitives to

More information

Machine Learning (BSMC-GA 4439) Wenke Liu

Machine Learning (BSMC-GA 4439) Wenke Liu Machine Learning (BSMC-GA 4439) Wenke Liu 01-25-2018 Outline Background Defining proximity Clustering methods Determining number of clusters Other approaches Cluster analysis as unsupervised Learning Unsupervised

More information

IMP: A Message-Passing Algorithm for Matrix Completion

IMP: A Message-Passing Algorithm for Matrix Completion IMP: A Message-Passing Algorithm for Matrix Completion Henry D. Pfister (joint with Byung-Hak Kim and Arvind Yedla) Electrical and Computer Engineering Texas A&M University ISTC 2010 Brest, France September

More information

Collaborative Filtering and Evaluation of Recommender Systems

Collaborative Filtering and Evaluation of Recommender Systems Collaborative Filtering and Evaluation of Recommender Systems Francesco Ricci ecommerce and Tourism Research Laboratory Automated Reasoning Systems Division ITC-irst Trento Italy ricci@itc.it http://ectrl.itc.it

More information

2nd Year Computational Physics Week 1 (experienced): Series, sequences & matrices

2nd Year Computational Physics Week 1 (experienced): Series, sequences & matrices 2nd Year Computational Physics Week 1 (experienced): Series, sequences & matrices 1 Last compiled September 28, 2017 2 Contents 1 Introduction 5 2 Prelab Questions 6 3 Quick check of your skills 9 3.1

More information

COMP 558 lecture 19 Nov. 17, 2010

COMP 558 lecture 19 Nov. 17, 2010 COMP 558 lecture 9 Nov. 7, 2 Camera calibration To estimate the geometry of 3D scenes, it helps to know the camera parameters, both external and internal. The problem of finding all these parameters is

More information

Geometric optics. The University of Texas at Austin CS384G Computer Graphics Don Fussell

Geometric optics. The University of Texas at Austin CS384G Computer Graphics Don Fussell Ray Tracing Geometric optics Modern theories of light treat it as both a wave and a particle. We will take a combined and somewhat simpler view of light the view of geometric optics. Here are the rules

More information

Algorithms. All-Pairs Shortest Paths. Dong Kyue Kim Hanyang University

Algorithms. All-Pairs Shortest Paths. Dong Kyue Kim Hanyang University Algorithms All-Pairs Shortest Paths Dong Kyue Kim Hanyang University dqkim@hanyang.ac.kr Contents Using single source shortest path algorithms Presents O(V 4 )-time algorithm, O(V 3 log V)-time algorithm,

More information

CS612 - Algorithms in Bioinformatics

CS612 - Algorithms in Bioinformatics Fall 2017 Structural Manipulation November 22, 2017 Rapid Structural Analysis Methods Emergence of large structural databases which do not allow manual (visual) analysis and require efficient 3-D search

More information

CSC-140 Assignment 5

CSC-140 Assignment 5 CSC-140 Assignment 5 Please do not Google a solution to these problems, cause that won t teach you anything about programming - the only way to get good at it, and understand it, is to do it! 1 Introduction

More information

Slide credit from Hung-Yi Lee & Richard Socher

Slide credit from Hung-Yi Lee & Richard Socher Slide credit from Hung-Yi Lee & Richard Socher 1 Review Word Vector 2 Word2Vec Variants Skip-gram: predicting surrounding words given the target word (Mikolov+, 2013) CBOW (continuous bag-of-words): predicting

More information

Implementing a Speech Recognition System on a GPU using CUDA. Presented by Omid Talakoub Astrid Yi

Implementing a Speech Recognition System on a GPU using CUDA. Presented by Omid Talakoub Astrid Yi Implementing a Speech Recognition System on a GPU using CUDA Presented by Omid Talakoub Astrid Yi Outline Background Motivation Speech recognition algorithm Implementation steps GPU implementation strategies

More information

3D Viewing. CS 4620 Lecture 8

3D Viewing. CS 4620 Lecture 8 3D Viewing CS 46 Lecture 8 13 Steve Marschner 1 Viewing, backward and forward So far have used the backward approach to viewing start from pixel ask what part of scene projects to pixel explicitly construct

More information

Searching and mining sequential data

Searching and mining sequential data Searching and mining sequential data! Panagiotis Papapetrou Professor, Stockholm University Adjunct Professor, Aalto University Disclaimer: some of the images in slides 62-69 have been taken from UCR and

More information

Programming Exercise 4: Neural Networks Learning

Programming Exercise 4: Neural Networks Learning Programming Exercise 4: Neural Networks Learning Machine Learning Introduction In this exercise, you will implement the backpropagation algorithm for neural networks and apply it to the task of hand-written

More information

CMPT 882 Week 3 Summary

CMPT 882 Week 3 Summary CMPT 882 Week 3 Summary! Artificial Neural Networks (ANNs) are networks of interconnected simple units that are based on a greatly simplified model of the brain. ANNs are useful learning tools by being

More information

Hypercubes. (Chapter Nine)

Hypercubes. (Chapter Nine) Hypercubes (Chapter Nine) Mesh Shortcomings: Due to its simplicity and regular structure, the mesh is attractive, both theoretically and practically. A problem with the mesh is that movement of data is

More information

CS 787: Assignment 4, Stereo Vision: Block Matching and Dynamic Programming Due: 12:00noon, Fri. Mar. 30, 2007.

CS 787: Assignment 4, Stereo Vision: Block Matching and Dynamic Programming Due: 12:00noon, Fri. Mar. 30, 2007. CS 787: Assignment 4, Stereo Vision: Block Matching and Dynamic Programming Due: 12:00noon, Fri. Mar. 30, 2007. In this assignment you will implement and test some simple stereo algorithms discussed in

More information

Embeddability of Arrangements of Pseudocircles into the Sphere

Embeddability of Arrangements of Pseudocircles into the Sphere Embeddability of Arrangements of Pseudocircles into the Sphere Ronald Ortner Department Mathematik und Informationstechnologie, Montanuniversität Leoben, Franz-Josef-Straße 18, 8700-Leoben, Austria Abstract

More information

Temporal Decision Trees for Diagnosis

Temporal Decision Trees for Diagnosis Dipartimento di Informatica Model-Based Diagnosis Model = description of a system to be diagnosed objective description. usually component oriented. may describe only correct behaviour or both correct

More information

MapReduce and Hadoop. Debapriyo Majumdar Indian Statistical Institute Kolkata

MapReduce and Hadoop. Debapriyo Majumdar Indian Statistical Institute Kolkata MapReduce and Hadoop Debapriyo Majumdar Indian Statistical Institute Kolkata debapriyo@isical.ac.in Let s keep the intro short Modern data mining: process immense amount of data quickly Exploit parallelism

More information

CPSC 340: Machine Learning and Data Mining. Principal Component Analysis Fall 2017

CPSC 340: Machine Learning and Data Mining. Principal Component Analysis Fall 2017 CPSC 340: Machine Learning and Data Mining Principal Component Analysis Fall 2017 Assignment 3: 2 late days to hand in tonight. Admin Assignment 4: Due Friday of next week. Last Time: MAP Estimation MAP

More information

Motion Detection. Final project by. Neta Sokolovsky

Motion Detection. Final project by. Neta Sokolovsky Motion Detection Final project by Neta Sokolovsky Introduction The goal of this project is to recognize a motion of objects found in the two given images. This functionality is useful in the video processing

More information

Gesture Recognition using Neural Networks

Gesture Recognition using Neural Networks Gesture Recognition using Neural Networks Jeremy Smith Department of Computer Science George Mason University Fairfax, VA Email: jsmitq@masonlive.gmu.edu ABSTRACT A gesture recognition method for body

More information

Computer Graphics. Lecture 14 Bump-mapping, Global Illumination (1)

Computer Graphics. Lecture 14 Bump-mapping, Global Illumination (1) Computer Graphics Lecture 14 Bump-mapping, Global Illumination (1) Today - Bump mapping - Displacement mapping - Global Illumination Radiosity Bump Mapping - A method to increase the realism of 3D objects

More information

Data Partitioning. Figure 1-31: Communication Topologies. Regular Partitions

Data Partitioning. Figure 1-31: Communication Topologies. Regular Partitions Data In single-program multiple-data (SPMD) parallel programs, global data is partitioned, with a portion of the data assigned to each processing node. Issues relevant to choosing a partitioning strategy

More information

Lecture 16: Introduction to Dynamic Programming Steven Skiena. Department of Computer Science State University of New York Stony Brook, NY

Lecture 16: Introduction to Dynamic Programming Steven Skiena. Department of Computer Science State University of New York Stony Brook, NY Lecture 16: Introduction to Dynamic Programming Steven Skiena Department of Computer Science State University of New York Stony Brook, NY 11794 4400 http://www.cs.sunysb.edu/ skiena Problem of the Day

More information

Towards direct motion and shape parameter recovery from image sequences. Stephen Benoit. Ph.D. Thesis Presentation September 25, 2003

Towards direct motion and shape parameter recovery from image sequences. Stephen Benoit. Ph.D. Thesis Presentation September 25, 2003 Towards direct motion and shape parameter recovery from image sequences Stephen Benoit Ph.D. Thesis Presentation September 25, 2003 September 25, 2003 Towards direct motion and shape parameter recovery

More information

Missing trace interpolation and its enhancement of seismic processes

Missing trace interpolation and its enhancement of seismic processes Missing trace interpolation Missing trace interpolation and its enhancement of seismic processes Wai-kin Chan and Robert R. Stewart ABSTRACT Many multi-channel seismic algorithms assume that the input

More information

Chapter. 9-1 Using the System Settings Menu 9-2 Memory Operations 9-3 System Settings 9-4 Reset 9-5 Tutorial Lock (ALGEBRA FX 2.

Chapter. 9-1 Using the System Settings Menu 9-2 Memory Operations 9-3 System Settings 9-4 Reset 9-5 Tutorial Lock (ALGEBRA FX 2. Chapter System Settings Menu Use the system settings menu to view system information and make system settings. The system settings menu lets you do the following. View memory usage information Make contrast

More information

Extended target tracking using PHD filters

Extended target tracking using PHD filters Ulm University 2014 01 29 1(35) With applications to video data and laser range data Division of Automatic Control Department of Electrical Engineering Linöping university Linöping, Sweden Presentation

More information

SLANG Session 4. Jason Quinley Roland Mühlenbernd Seminar für Sprachwissenschaft University of Tübingen

SLANG Session 4. Jason Quinley Roland Mühlenbernd Seminar für Sprachwissenschaft University of Tübingen SLANG Session 4 Jason Quinley Roland Mühlenbernd Seminar für Sprachwissenschaft University of Tübingen Overview Network properties Degree Density and Distribution Clustering and Connections Network formation

More information