L. Zhao's scientific contributions

... The deficiencies such as small convergence rate [4], overfitting problem [5], of classical gradient based methods lead to the development of variant of neural network, i.e., extreme learning machine (ELM) introduced by Huang [6]. It is employed for sales forecasting [7], bankruptcy prediction [8], stock price prediction [2], electricity price forecasting [9], temperature prediction of molten steel [10] and predicting patient-outcomes [11] etc. A revolutionary ELM has many benefits of improving training accuracy [12], arbitrary selection of weights of input layers with biases [13], preventing from local optimum [14,15], resolving over-fitting problem [16], rapid prediction in non-linear systems occurred in various real-life situations [2,4,17], requiring less human involvement [18]. The ELM associates with the problems such as at every trial, the prediction accuracy result of ELM varies with their arbitrary association between the input layer and hidden layer [19,20], requires more number of hidden layer nodes increasing the training time [12,21], fails to provide satisfactory result in few cases due to the poor convergence [3]. ...