Going Deeper with CKELM
Yang Fang, Wenxin Hu · 2017
Convolutional Neural Networks (CNN) is known for its features extraction from images. Extreme Learning Machine (ELM), has been adopted in various researching fields for its fast learning. CNN-KELM has overcome the KELM'S shortage of extracting features as this model uses CNN to detect the input image before classifying with KELM. Our improved CKELM includes rank based pooling strategy and GPU training. We make experiments to evaluate our proposal on dataset CIFAR and result proves that our pooling methods perform better than traditional ones.