104
Voted
ICALP
2009
Springer
16 years 21 days 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....