it is not about the subject it is about data, you must first look at your data , which type ,is it :'big data or small data', is it:time series or multivariate data , is it : images or signals, do you want to implement the algoritm in real time or just you need data to train and test the learning model? i will give you here some sugestions:
1)if your data is driven seuentialy in time, you must use online sequential ELM.
2)if your data is big i advice you to use deep belief ELM.
3) if 1 and 2 happened in the same time then you have to use online sequential deep belief nets based ELM.
4) if your data are images you beter be used Convolotionel neural net based ELM.
5) if your data are multivaraite signals you ELM according to 1 2 3;
if you didnt found your type of data here , contact me, you will found my email on may profile,
and if the answer is helpful , accept it and thanks