Challenges in representation learning: a report on three machine learning contests.

The ICML 2013 Workshop on Challenges in Representation Learning(1) focused on three challenges: the black box learning challenge, the facial expression recognition challenge, and the multimodal learning challenge. We describe the datasets created for these challenges and summarize the results of the competitions. We provide suggestions for organizers of future challenges and some comments on what kind of knowledge can be gained from machine learning competitions.

Mutual passivation of donors and isovalent nitrogen in GaAs.

We study the mutual passivation of shallow donor and isovalent N in GaAs. We find that all the donor impurities, SiGa, GeGa, SAs, and SeAs, bind to N in GaAs:N, which has a large N-induced band-gap reduction relative to GaAs. For a group-IV impurity such as Si, the formation of the nearest-neighbor SiGa-NAs defect complex creates a deep donor level below the conduction band minimum (CBM). The coupling between this defect level with the CBM pushes the CBM upwards, thus restoring the GaAs band gap; the lowering of the defect level relative to the isolated SiGa shallow donor level is responsible for the increased electrical resistivity. Therefore, Si and N mutually passivate each other's electrical and optical activities in GaAs. For a group-VI shallow donor such as S, the binding between SAs and NAsdoes not form a direct bond; therefore, no mutual passivation exists in the GaAs:(S+N) system.