联合算法稳定下的排一推广误差界

The Bound of Leave-one-out Generalization Error under United Stability

在机器学习算法的联合稳定性条件下,利用广义McDiarmid不等式获得了排一推广误差的界。讨论了机器学习的推广能力,和类似的结果进行了比较。

Under the assumption of united stability, a bound of leave-one-out generalization error is derived by using extended versions of McDiarmid＇s inequality, the generalization ability of machine learning is also discussed and compared to similar results.