In this paper we present our findings on how representation learning on large unlabeled speech corpora can be beneficially utilized for speech emotion recognition (SER). Prior work on representation ...