Time Series Classification¶
Perform time series classification in Amazon SageMaker
SageMaker Algorithm |
CPU Training |
GPU Training |
Multi-GPU Training |
Incremental Training |
|---|---|---|---|---|
LSTM-FCN SageMaker Algorithm¶
The LSTM-FCN SageMaker Algorithm performs time series classification with the Long Short-Term Memory Fully Convolutional Network (LSTM-FCN). The LSTM-FCN model consists of two blocks: a recurrent block and a convolutional block. The two blocks process the input time series in parallel. After that their output representations are concatenated and passed to a final linear layer. The algorithm can be used for binary, multiclass and multilabel classification of univariate time series. For additional information, see the algorithm's AWS Marketplace listing page and GitHub repository.
InceptionTime SageMaker Algorithm¶
The InceptionTime SageMaker Algorithm performs time series classification with the InceptionTime Network. The InceptionTime network consists of a stack of Inception blocks linked by residual connections. Each block processes the inputs in parallel using three convolutional layers and a max pooling layer and returns a combined representation. The algorithm trains an ensemble of InceptionTime networks and generates the final predicted class labels by averaging the class probabilities predicted by the different networks in the ensemble. The algorithm can be used for binary, multiclass and multilabel classification of both univariate and multivariate time series. For additional information, see the algorithm's AWS Marketplace listing page and GitHub repository.