英语翻译The key difference to Theorem 1 is that the regret bound of the exponentiated algorithm scales logarith- mically with the number of features,and with the `1-norm of w,which can be advantageous if the optimal w is sparse.

英语翻译The key difference to Theorem 1 is that the regret bound of the exponentiated algorithm scales logarith- mically with the number of features,and with the `1-norm of w,which can be advantageous if the optimal w is sparse.
英语翻译
The key difference to Theorem 1 is that the regret bound of the exponentiated algorithm scales logarith- mically with the number of features,and with the `1-norm of w,which can be advantageous if the optimal w is sparse.

英语翻译The key difference to Theorem 1 is that the regret bound of the exponentiated algorithm scales logarith- mically with the number of features,and with the `1-norm of w,which can be advantageous if the optimal w is sparse.
定理1的主要区别是,绑定的遗憾幂算法级对数mically的特点,并与1标准的W,这可能是有利的,如果的最佳w是稀疏的.

关键地ff分定理1是遗憾的幂算法结合了尺度对数与功能的数量，和W的`模，可有利地最佳W是稀疏的。

关键的区别是,后悔定理1绑定的exponentiated算法——mically logarith鳞片与数量的特性,并以‘1规范的w,可以有利如果最优w是稀疏的。