Most existing Zero-Shot Learning (ZSL) methods have the strong bias problem, in which instances of unseen (target) classes tend to be categorized as one of the seen (source) classes. So they yield poor performance after being deployed in the generalized ZSL settings. In this paper, we propose a straightforward yet effective method named Quasi-Fully Supervised Learning (QFSL) to alleviate the bias problem. Our method follows the way of transductive learning, which assumes that both the labeled source images and unlabeled target images are available for training. In the semantic embedding space, the labeled source images are mapped to several fixed points specified by the source categories, and the unlabeled target images are forced to be mapped to other points specified by the target categories. Experiments conducted on AwA2, CUB and SUN datasets demonstrate that our method outperforms existing state-ofthe-art approaches by a huge margin of 9.3 ∼ 24.5% following generalized ZSL settings, and by a large margin of 0.2 ∼ 16.2% following conventional ZSL settings.

No abstract

We have synthesized a new oxypnictide, Ba2Ti2Fe2As4O, via a solid-state reaction under a vacuum. The compound crystallizes in a body-centered tetragonal lattice, which can be viewed as an intergrowth of BaFe2As2 and BaTi2As2O, thus containing Fe2As2 layers and Ti2O sheets. Bulk superconductivity at 21 K is observed after annealing the as-prepared sample at 773 K for 40 h. In addition, an anomaly in resistivity and magnetic susceptibility around 125 K is revealed, suggesting a charge- or spin-density wave transition in the Ti sublattice.

Knowledge Distillation (KD) has made remarkable progress in the last few years and become a popular paradigm for model compression and knowledge transfer. However, almost all existing KD algorithms are datadriven, i.e., relying on a large amount of original training data or alternative data, which is usually unavailable in real-world scenarios. In this paper, we devote ourselves to this challenging problem and propose a novel adversarial distillation mechanism to craft a compact student model without any real-world data. We introduce a model discrepancy to quantificationally measure the difference between student and teacher models and construct an optimizable upper bound. In our work, the student and the teacher jointly act the role of the discriminator to reduce this discrepancy, when a generator adversarially produces some "hard samples" to enlarge it. Extensive experiments demonstrate that the proposed data-free method yields comparable performance to existing data-driven methods. More strikingly, our approach can be directly extended to semantic segmentation, which is more complicated than classification and our approach achieves state-of-the-art results. The code will be released.

Motivated by the rich interplay among electronic correlation, spin-orbit coupling (SOC), crystal-field splitting, and geometric frustrations in the honeycomblike lattice, we systematically investigated the electronic and magnetic properties of Li 2 RhO 3 . The material is semiconducting with a narrow band gap of ∼ 78 meV, and its temperature dependence of resistivity conforms to a three-dimensional variable range hopping mechanism. No long-range magnetic ordering was found down to 0.5 K, due to the geometric frustrations. Instead, single atomic spin-glass behavior below the spin-freezing temperature (∼6 K) was observed and its spin dynamics obeys the universal critical slowing down scaling law. A first-principles calculation suggested it to be a relativistic Mott insulator mediated by both electronic correlation and SOC. With moderate strength of electronic correlation and SOC, our results shed light on the research of the Heisenberg-Kitaev model in realistic materials.