Random Convolution

Random convolution methods leverage randomly generated convolutional kernels to efficiently extract features from time series data, primarily aiming to improve the speed and accuracy of time series classification. Current research focuses on optimizing existing architectures like ROCKET and MiniRocket, including techniques like feature selection and pruning to reduce computational cost while maintaining or improving accuracy, and extending these methods to multivariate time series and other applications such as dynamic functional connectivity analysis. These advancements offer significant potential for improving the scalability and interpretability of time series analysis across diverse scientific fields and practical applications, such as healthcare, finance, and environmental monitoring.

Papers

September 2, 2024

Time series classification with random convolution kernels based transforms: pooling operators and input representations matter
Mouhamadou Mansour Lo, Gildas Morvan, Mathieu Rossi, Fabrice Morganti, David Mercier
Time Series Classification Feature Operator Input Representation Random Convolution

August 5, 2024

Classification of Raw MEG/EEG Data with Detach-Rocket Ensemble: An Improved ROCKET Algorithm for Multivariate Time Series Analysis
Adrià Solana, Erik Fransén, Gonzalo Uribarri
Multivariate Time Series EEG Data Multivariate Time Series Classification Synthetic Time Series EEG Datasets Brain Data Random Convolution Magnetoencephalography Data

June 24, 2024

Generalized Dynamic Brain Functional Connectivity Based on Random Convolutions
Yongjie Duan, Vince D. Calhoun, Zhiying Long
LeArning Abstract Sliding Window Random Convolution Dynamic Functional Connectivity Connectivity Matrix

April 28, 2024

Evaluating ROCKET and Catch22 features for calf behaviour classification from accelerometer data using Machine Learning models
Oshana Dissanayake, Sarah E. McPherson, Joseph Allyndree, Emer Kennedy, Padraig Cunningham, Lucile Riaboff
Machine Learning Model Challenge Task Accelerometer Data Precision Livestock Farming Random Convolution

November 16, 2023

Polynomially Over-Parameterized Convolutional Neural Networks Contain Structured Strong Winning Lottery Tickets
Arthur da Cunha, Francesco d'Amore, Emanuele Natale
Convolution Layer Structured Pruning Estimated Team Strength Lottery Ticket Lottery Ticket Hypothesis Unstructured Pruning Random Convolution Randomly Initialized

September 25, 2023

Detach-ROCKET: Sequential feature selection for time series classification with random convolutional kernels
Gonzalo Uribarri, Federico Barone, Alessio Ansuini, Erik Fransén
Recurrent Neural Network Time Series Data Time Series Classification Random Convolution

September 15, 2023

POCKET: Pruning Random Convolution Kernels for Time Series Classification from a Feature Selection Perspective
Shaowu Chen, Weize Sun, Lei Huang, Xiaopeng Li, Qingyuan Wang, Deepu John
Time Series Classification Random Convolution Sequential Classifier New POCKET

May 17, 2023

Time Series Clustering With Random Convolutional Kernels
Jorge Marco-Blanco, Rubén Cuevas
Time Series Clustering Random Convolution Time Series Data Mining

April 2, 2023

Progressive Random Convolutions for Single Domain Generalization
Seokeon Choi, Debasmit Das, Sungha Choi, Seunghan Yang, Hyunsin Park, Sungrack Yun
Single Domain Generalization Random Convolution

January 26, 2023

The Power of Linear Combinations: Learning with Random Convolutions
Paul Gavrikov, Janis Keuper
Convolutional Neural Network Real Power CNN Architecture Convolutional Filter Linear Combination Convolution Based Random Convolution

April 4, 2022

Taking ROCKET on an Efficiency Mission: Multivariate Time Series Classification with LightWaveS
Leonardos Pantiskas, Kees Verstoep, Mark Hoogendoorn, Henri Bal
Enhancing Efficiency Multivariate Time Series Classification Prediction Speed Random Convolution

March 7, 2022

S-Rocket: Selective Random Convolution Kernels for Time Series Classification
Hojjat Salehinejad, Yang Wang, Yuanhao Yu, Tang Jin, Shahrokh Valaee
Time Series Classification Ridge Regression Convolutional Kernel Convolution Operation Random Convolution Kernel Trick

Random Convolution

Papers

Time series classification with random convolution kernels based transforms: pooling operators and input representations matter

Classification of Raw MEG/EEG Data with Detach-Rocket Ensemble: An Improved ROCKET Algorithm for Multivariate Time Series Analysis

Generalized Dynamic Brain Functional Connectivity Based on Random Convolutions

Evaluating ROCKET and Catch22 features for calf behaviour classification from accelerometer data using Machine Learning models

Polynomially Over-Parameterized Convolutional Neural Networks Contain Structured Strong Winning Lottery Tickets

Detach-ROCKET: Sequential feature selection for time series classification with random convolutional kernels

POCKET: Pruning Random Convolution Kernels for Time Series Classification from a Feature Selection Perspective

Time Series Clustering With Random Convolutional Kernels

Progressive Random Convolutions for Single Domain Generalization

The Power of Linear Combinations: Learning with Random Convolutions

Taking ROCKET on an Efficiency Mission: Multivariate Time Series Classification with LightWaveS

S-Rocket: Selective Random Convolution Kernels for Time Series Classification