Show me your friends, I'll tell you your emotions: Emotional fit of immigrant-origin minority youth in cross-cultural friendship networks.

The typical emotional responses to certain types of situations differ across cultures. Being reprimanded by your teacher in front of the class may be cause for anger and indignation among pupils in one cultural context, but for anger, shame, and possibly respect for the teacher among pupils in another cultural context. The consequence for immigrant-origin minorities is that they may not fit the emotions of the majority culture. Previous research has found that minorities who have majority contact have higher emotional fit with the majority culture. In the current study, we suggest that friendships with majority peers are particularly important to minorities' emotional fit. Students (945 minority and 1256 majority) from a representative sample of Belgian middle schools completed a sociometric questionnaire on their classroom friendships and rated their emotional experiences in two situations. Multilevel models yielded higher levels of emotional fit for minority youth with many (vs. few) majority friends as well as for minorities whose majority friends are connected (vs. less connected) to each other, or who are well-connected in the majority peer network. Having majority friends predicted emotional fit over and above majority contact in general.

Training recurrent networks by Evolino.

In recent years, gradient-based LSTM recurrent neural networks (RNNs) solved many previously RNN-unlearnable tasks. Sometimes, however, gradient information is of little use for training RNNs, due to numerous local minima. For such cases, we present a novel method: EVOlution of systems with LINear Outputs (Evolino). Evolino evolves weights to the nonlinear, hidden nodes of RNNs while computing optimal linear mappings from hidden state to output, using methods such as pseudo-inverse-based linear regression. If we instead use quadratic programming to maximize the margin, we obtain the first evolutionary recurrent support vector machines. We show that Evolino-based LSTM can solve tasks that Echo State nets (Jaeger, 2004a) cannot and achieves higher accuracy in certain continuous function generation tasks than conventional gradient descent RNNs, including gradient-based LSTM.