Algorithm Description

 

The provided class labels for the 3 test files are derived from precomputed features.

 

We only focus on the training and testing of the classifier: single layer feedforward neural network (SLFN) based on a recently proposed learning algorithm: Extreme Learning Machine (ELM)[1]. The simulation includes 4 steps:

 

1, Model Selection: The optimal number of hidden neurons of SLFN is estimated by train and validation set approach.

2, Training Phase: The weights and biases of SLFN are estimated by ELM using training data.

3, Testing Phase: The testing output vectors are derived by the trained SLFN for each testing data.

4, Output vector fusion and 8-smple combination: An averaging filter of 80-sample window is applied to the testing output vectors and followed by Vote strategy. Vote strategy is deployed to produce the class labels for each group of 8 averaged testing output vectors every 0.5 seconds.

 

Reference

l         G.-B. Huang, Q.-Y. Zhu, and C.-K. Siew, “Extreme Learning Machine: A New Learning Scheme of Feedforward Neural Networks”, 2004 International Joint Conference on Neural Networks (IJCNN'2004), July 25-29, 2004, Budapest, Hungary.



[1] Introduction and source code of ELM can be found in: http://www.ntu.edu.sg/home/egbhuang/