Learning using multidimensional internal rewards | IEEE Conference Publication | IEEE Xplore