1 Erratum to: Neural Comput & Applic DOI 10.1007/s00521-015-1909-9

In the original publication, some of the equations and Fig. 6 have been cited incorrectly.

In Sect. 3.2, paragraph four, the article states:

“In Eq. 4, β is the matrix of output weight and Y is the matrix of class label, which can be expressed, respectively, as”

Correction: This should read as, “In Eq. 5, β is the matrix of output weight and Y is the matrix of class label, which can be expressed, respectively, as”

In Sect. 3.2, paragraph five, the article states:

“Therefore, the training process of ELM is equivalent to solve the linear Eq. 4, and the output weights β can be estimated”

Correction: This text should refer Eq. 5 as “Therefore, the training process of ELM is equivalent to solve the linear Eq. 5, and the output weights β can be estimated”

In Sect. 3.2, paragraph 6, the article states:

“It is worth noting that because the RTSSP is a binary classification problem, we consider there is only one node in the output layer, and the output dimension m is set to 1 in Eq. 8 and hereafter.”

Correction: This text should refer Eq. 9 as, “It is worth noting that because the RTSSP is a binary classification problem, we consider there is only one node in the output layer, and the output dimension m is set to 1 in Eq. 9 and hereafter.”

In Sect. 3.3, paragraph two, the article states:

“The Lagrangian function associated with the equality constrained problem of Eq. 9 is”

Correction: This should read as “The Lagrangian function associated with the equality constrained problem of Eq. 10 is”

In Sect. 5.3, paragraph two, the article states:

As can be seen in Fig. 6, cost-sensitive methods are more suitable than cost-blind methods since they achieve lower Costs and FDR.

Correction: This should refer to Fig. 7.