Skip to Main Content
We describe a learning algorithm for the prediction of splice site locations in human DNA in the presence of sequence annotation errors in the training data. Experimental results on a common dataset including errors are reported. We also give an efficient implementation. The resulting software package is publicly available.