ICALP
2009
Springer
14 years 11 months ago
2009
Springer
We give new algorithms for learning halfspaces in the challenging malicious noise model, where an adversary may corrupt both the labels and the underlying distribution of examples....