Tsai
시계열 및 시퀀스를 위한 오픈소스 딥러닝 라이브러리
Time series Timeseries Deep Learning Machine Learning Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai
Features
- Pytorch 와 fastai 기반
- classification, regression, forecasting, imputation 같은 시계열 작업에 중점
Models
사용 가능한 State-of-the-art 모델들
- LSTM (Hochreiter, 1997)
- GRU (Cho, 2014)
- MLP - Multilayer Perceptron (Wang, 2016)
- FCN - Fully Convolutional Network (Wang, 2016)
- ResNet - Residual Network (Wang, 2016)
- LSTM-FCN (Karim, 2017)
- GRU-FCN (Elsayed, 2018)
- mWDN - Multilevel wavelet decomposition network (Wang, 2018)
- TCN - Temporal Convolutional Network (Bai, 2018)
- MLSTM-FCN - Multivariate LSTM-FCN (Karim, 2019)
- InceptionTime (Fawaz, 2019)
- Rocket (Dempster, 2019)
- XceptionTime (Rahimian, 2019)
- ResCNN - 1D-ResCNN (Zou , 2019)
- TabModel - modified from fastai’s TabularModel
- OmniScale - Omni-Scale 1D-CNN (Tang, 2020)
- TST - Time Series Transformer (Zerveas, 2020)
- TabTransformer (Huang, 2020)
- MiniRocket (Dempster, 2021)
- XCM - An Explainable Convolutional Neural Network (Fauvel, 2021)
- gMLP - Gated Multilayer Perceptron (Liu, 2021)
- GatedTabTransformer (Cholakov, 2022)