Magnetic reconnection is a fundamental process at work in laboratory, space and astrophysical plasmas, in which magnetic field lines change their topology and convert magnetic energy to plasma particles by acceleration and heating. One of the most important problems in reconnection research has been to understand why reconnection occurs so much faster than predicted by MHD theory. Following the recent pedagogical review of this subject [M. Yamada, R. Kulsrud, and H. Ji, Rev. Mod. Phys. {\bf 82}, 603 (2010)], this paper presents a review of more recent discoveries and findings in the research of fast magnetic reconnection in laboratory, space, and astrophysical plasmas. In spite of the huge difference in physical scales, we find remarkable commonality between the characteristics of the magnetic reconnection in laboratory and space plasmas. In this paper, we will focus especially on the energy flow, a key feature of the reconnection process. In particular the experimental results on the energy conversion and partitioning in a laboratory reconnection layer [M. Yamada {\it et al.}, Nat. Commu. {\bf 5}, 4474 (2014)] are discussed and compared with quantitative estimates based on two-fluid analysis. In the Magnetic Reconnection Experiment (MRX), we find that energy deposition to electrons is localized near the X-point and is mostly from the electric field component perpendicular to the magnetic field. The mechanisms of ion acceleration and heating are also identified and a systematic and quantitative study on the inventory of converted energy within a reconnection layer with a well-defined but variable boundary. The measured energy partition in a reconnection region of similar effective size ($L \approx$ 3 ion skin depths) of the Earth's magneto-tail [J. Eastwood {\it et al.}, Phys. Rev. Lett. {\bf 110}, 225001 (2013)] is notably consistent with our laboratory results. Finally, to study the global aspects of magnetic reconnection, we have carried out a laboratory experiment on the stability criteria for solar flare eruptions, including {\textquotedblleft}storage and release{\textquotedblright} mechanisms of magnetic energy. We show that toroidal magnetic flux generated by magnetic relaxation (reconnection) processes generates a new stabilizing force which prevents plasma eruption. This result has lead us to discovery of a new stabilizing force for solar flares [C. E. Myers {\it et al.}, Nature {\bf 528}, 526 (2015)]
Extrapolation -- the ability to make inferences that go beyond the scope of one's experiences -- is a hallmark of human intelligence. By contrast, the generalization exhibited by contemporary neural network algorithms is largely limited to interpolation between data points in their training corpora. In this paper, we consider the challenge of learning representations that support extrapolation. We introduce a novel visual analogy benchmark that allows the graded evaluation of extrapolation as a function of distance from the convex domain defined by the training data. We also introduce a simple technique, context normalization, that encourages representations that emphasize the relations between objects. We find that this technique enables a significant improvement in the ability to extrapolate, considerably outperforming a number of competitive techniques.
Gilson, Erik; Lee, H.; Bortolon, A.; Choe, W.; Diallo, A.; Hong, S. H.; Lee, H. M.; Maingi, R.; Mansfield, D. K.; Nagy, A.; Park, S. H.; Song, I. W.; Song, J. I.; Yun, S. W.; Yoon, S. W.; Nazikian, R.
