Convolutional neural networks have gained a remarkable success in computer vision. However, most usable network architectures are hand-crafted and usually require expertise and elaborate design. In this paper, we provide a block-wise network generation pipeline called BlockQNN which automatically builds high-performance networks using the Q-Learning paradigm with epsilon-greedy exploration strategy. The optimal network block is constructed by the learning agent which is trained sequentially to choose component layers. We stack the block to construct the whole auto-generated network. To accelerate the generation process, we also propose a distributed asynchronous framework and an early stop strategy. The block-wise generation brings unique advantages: (1) it performs competitive results in comparison to the hand-crafted state-of-the-art networks on image classification, additionally, the best network generated by BlockQNN achieves 3.54% top-1 error rate on CIFAR-10 which beats all existing auto-generate networks.(2) in the meanwhile, it offers tremendous reduction of the search space in designing networks which only spends 3 days with 32 GPUs, and (3) moreover, it has strong generalizability that the network built on CIFAR also performs well on a larger-scale ImageNet dataset. * This work was done when Zhao Zhong worked as an intern at Sense-Time Research. performance network architecture generally possesses a tremendous number of possible configurations about the number of layers, hyperparameters in each layer and type of each layer. It is hence infeasible for manually exhaustive search, and the design of successful hand-crafted networks heavily rely on expert knowledge and experience. Therefore, constructing the network in a smart and automatic manner remains an open problem.Although some recent works have attempted computeraided or automated network design [2,37], there are several challenges still unsolved: (1) Modern neural networks always consist of hundreds of convolutional layers, each of which has numerous options in type and hyperparameters. It makes a huge search space and heavy computational costs for network generation. (2) One typically designed network is usually limited on a specific dataset or task, and thus is hard to transfer to other tasks or generalize to another dataset with different input data sizes. In this paper, we provide a solution to the aforementioned challenges by a novel fast Q-learning framework, called BlockQNN, to automatically design the network architecture, as shown in Fig. 1.Particularly, to make the network generation efficient and generalizable, we introduce the block-wise network generation, i.e., we construct the network architecture as a flexible stack of personalized blocks rather tedious per-layer network piling. Indeed, most modern CNN architectures such as Inception [30,14,31] and ResNet Series [10,11] are assembled as the stack of basic block structures. For example, the inception and residual blocks shown in Fig. 1 are repeatedly concatenated to construct the enti...
High T c superconductivity in FeAs-based (pnictides) multilayers, evading temperature decoherence effects in a quantum condensate, is assigned to a Feshbach resonance (called also shape resonance) in the exchange-like interband pairing. The resonance is switched on by tuning the chemical potential at an electronic topological transition (ETT) near a band edge, where the Fermi surface topology of one of the subbands changes from 1D to 2D topology. We show that the tuning is realized by changing i) the misfit strain between the superconducting planes and the spacers ii) the charge density and iii) the disorder. The system is at the verge of a catastrophe i.e. near a structural and magnetic phase transition associated with the stripes (analogous to the 1/8 stripe phase in cuprates) order to disorder phase transition. Fine tuning of both the chemical potential and the disorder pushes the critical temperature T s of this phase transition to zero giving a quantum critical point. Here the quantum lattice and magnetic fluctuations promote the Feshbach resonance of the exchange-like anisotropic pairing. This superconducting phase that resists to the attacks of temperature is shown to be controlled by the interplay of the hopping energy between stripes and the quantum fluctuations. The superconducting gaps in the multiple Fermi surface spots reported by the recent ARPES experiment of D. V. Evtushinsky et al. arXiv:0809.4455 are shown to support the Feshbach scenario.
Here we use global and local magnetometry and Hall probe imaging to investigate the electromagnetic connectivity of the superconducting current path in the oxygendeficient fluorine-free Nd-based oxypnictides. High resolution transmission electron microscopy and scanning electron microscopy show strongly-layered crystallites, evidence for a ~ 5nm amorphous oxide around individual particles, and second phase neodymium oxide which may be responsible for the large paramagnetic background at high field and at high temperatures. From global magnetometry and electrical transport measurements it is clear that there is a small supercurrent flowing on macroscopic sample dimensions (mm), with a lower bound for the average (over this length scale) critical current density of the order of 10 3 A/cm 2 . From magnetometry of powder samples and local Hall probe imaging of a single large conglomerate particle ~120 microns it is clear that on smaller scales, there is better current connectivity with a critical current density of the order of 5 x 10 4 A/cm 2 . We find enhanced flux creep around the second peak anomaly in the magnetisation curve and an irreversibility line significantly below H c2 (T) as determined by ac calorimetry.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.