No abstract

Neural Architecture Search (NAS) is a promising and rapidly evolving research area. Training a large number of neural networks requires an exceptional amount of computational power, which makes NAS unreachable for those researchers who have limited or no access to high-performance clusters and supercomputers. A few benchmarks with precomputed neural architectures performances have been recently introduced to overcome this problem and ensure reproducible experiments. However, these benchmarks are only for the computer vision domain and, thus, are built from the image datasets and convolution-derived architectures. In this work, we step outside the computer vision domain by leveraging the language modeling task, which is the core of natural language processing (NLP). Our main contribution is as follows: we have provided search space of recurrent neural networks on the text datasets and trained 14k architectures within it; we have conducted both intrinsic and extrinsic evaluation of the trained models using datasets for semantic relatedness and language understanding evaluation; finally, we have tested several NAS algorithms to demonstrate how the precomputed results can be utilized. We consider that the benchmark will provide more reliable empirical findings in the community and stimulate progress in developing new NAS methods well suited for recurrent architectures.INDEX TERMS Benchmark, natural language processing, neural architecture search, recurrent neural network.

We further generalize the powerful method, which we have recently developed for description of the background matter influence on neutrinos, for the case of an electron moving in matter. On the basis of the modified Dirac equation for the electron, accounting for the standard model interaction with particles of the background, we predict and investigate in some detail a new mechanism of the electromagnetic radiation that is emitted by moving in matter electron due to its magnetic moment. We have termed this radiation the "spin light of electron" in matter and predicted that this radiation can have consequences accessible for experimental observations in astrophysical and cosmological settings.

We derive the modified Dirac equation for an electron undergos an influence of the standard model interaction with the nuclear matter. The exact solutions for this equation and the electron energy spectrum in matter are obtained. This establishes a rather powerful method for investigation of different processes that can appear when electrons propagate in background matter. On this basis we study in detail the spin light of electron in nuclear matter, a new type of electromagnetic radiation which can be emitted by an electron moving in dense matter.

Based on the method of exact solutions to the quantum equations for the wave functions of particles in external fields and media within the framework of the standard interaction model, the modified Dirac equation for the electron is derived that allows its interaction with the medium to be considered. An exact solution to the equation and energy spectrum of the electron states are determined. In the context of this approach, a new type of electromagnetic radiation -spin light of electron in a neutron medium -is predicted and studied. General expressions for the probability of the process in unit time and for the radiation intensity are derived, and a dependence of the radiation intensity on the electron energy and density of the medium is analyzed. The limiting cases of the process and polarization properties of radiation are investigated.