Figure 2: Training the samples by ELM or MAELM.