Ionic thermoelectrics show great potential in low-grade heat harvesting and thermal sensing owing to their ultrahigh thermopower, low cost and ease in production. However, the lack of effective n-type ionic thermoelectric materials seriously hinders
their applications. Here, we report giant and bidirectionally tunable thermopowers within an ultrawide range from -23 to +32 mV K-1 at 90% RH in solid ionic-liquid-based ionogels, rendering it among the best n- and p-type ionic thermoelectric materials. A novel thermopower regulation strategy through ion doping to selectively induce ion aggregates via strong ion-ion interactions is proposed. These charged aggregates are found decisive in modulating the sign and enlarging the magnitude of the thermopower in the ionogels. A prototype wearable device integrated with 12 p-n pairs is demonstrated with a total thermopower of 0.358 V K-1 in general indoor conditions, showing promise for ultrasensitive body heat detection.
Semantic segmentation is a basic but non-trivial task in computer vision. Many previous work focus on utilizing affinity patterns to enhance segmentation networks. Most of these studies use the affinity matrix as a kind of feature fusion weights, whi
ch is part of modules embedded in the network, such as attention models and non-local models. In this paper, we associate affinity matrix with labels, exploiting the affinity in a supervised way. Specifically, we utilize the label to generate a multi-scale label affinity matrix as a structural supervision, and we use a square root kernel to compute a non-local affinity matrix on output layers. With such two affinities, we define a novel loss called Affinity Regression loss (AR loss), which can be an auxiliary loss providing pair-wise similarity penalty. Our model is easy to train and adds little computational burden without run-time inference. Extensive experiments on NYUv2 dataset and Cityscapes dataset demonstrate that our proposed method is sufficient in promoting semantic segmentation networks.
We study a superconducting hetro-junction with one side characterized by the unconventional chiral $p$-wave gap function $p_xpm ip_y$ and the other side the conventional $s$-wave one. Though a relative phase of $pm frac{pi}{2}$ between any two compon
ents of gap functions is favored in the junction region, mutual phase differences cannot achieve $pm frac{pi}{2}$ simultaneously, which results in frustration. Based on a Ginzburg-Landau free energy analysis, the frustrated pattern is determined to be $s+ ieta_1 (e^{ ieta_2 varphi/2}p_x +eta_3 e^{- ieta_2 varphi/2}p_y)$ with $eta_j=pm 1$ ($j=1,2,3$), where $varphi$ is the phase difference between the $p_x$- and $p_y$-wave gap functions. Furthermore, we find that the junction exhibits an anisotropic magnetoelectric effect, manifesting itself as an anisotropic spin magnetization along the edge of the junction.
Recent experiments on the flow of helium-4 fluid through nanopores with tunable pore radius provide a platform for studying the quasi-one-dimensional (quasi-1D) superfluid behaviors. In the extreme 1D limit, the helium atoms are localized by disorder
ed small variations in the substrate potential provided by the pore walls. In the limit of wide pore radius, a solid layer of helium-4 is expected to coat the pore walls smoothing out the substrate potential, and superfluidity is observed in the central region. Building on earlier quantum Monte Carlo results, we propose a scenario for this crossover using a shell model of coupled Luttinger liquids. We find that a small radius pore will always localize the helium atoms, but above a critical radius, a single 1D channel flows through the pore and can be described by Luttinger liquid theory.
In this work, we perform a combination of classical and spin wave analysis on the one-dimensional spin-$S$ Kitaev-Heisenberg-Gamma model in the region of an antiferromagnetic Kitaev coupling. Four phases are found, including a Neel ordered phase, a p
hase with $O_hrightarrow D_3$ symmetry breaking, and $D_3$-breaking I, II phases which both break $D_3$ symmetries albeit in different ways, where $O_h$ is the full octahedral group and $D_3$ is the dihedral group of order six. The lowest-lying spin wave mass is calculated perturbatively in the vicinity of the hidden SU(2) symmetric ferromagnetic point.
A central question on Kitaev materials is the effects of additional couplings on the Kitaev model which is proposed to be a candidate for realizing topological quantum computations. However, two spatial dimension typically suffers the difficulty of l
acking controllable approaches. In this work, using a combination of powerful analytical and numerical methods available in one dimension, we perform a comprehensive study on the phase diagram of a one-dimensional version of the spin-1/2 Kitaev-Heisenberg-Gamma model in its full parameter space. A strikingly rich phase diagram is found with nine distinct phases, including four Luttinger liquid phases, a ferromagnetic phase, a Neel ordered phase, an ordered phase of distorted-spiral spin alignments, and two ordered phase which both break a $D_3$ symmetry albeit in different ways, where $D_3$ is the dihedral group of order six. Our work paves the way for studying one-dimensional Kitaev materials and may provide hints to the physics in higher dimensional situations.
A minimal Kitaev-Gamma model has been recently investigated to understand various Kitaev systems. In the one-dimensional Kitaev-Gamma chain, an emergent SU(2)$_1$ phase and a rank-1 spin ordered phase with $O_hrightarrow D_4$ symmetry breaking were i
dentified using non-Abelian bosonization and numerical techniques. However, puzzles near the antiferromagnetic Kitaev region with finite Gamma interaction remained unresolved. Here we focus on this parameter region and find that there are two new phases, namely, a rank-1 ordered phase with an $O_hrightarrow D_3$ symmetry breaking, and a peculiar Kitaev phase. Remarkably, the $O_hrightarrow D_3$ symmetry breaking corresponds to the classical magnetic order, but appears in a region very close to the antiferromagnetic Kitaev point where the quantum fluctuations are presumably very strong. In addition, a two-step symmetry breaking $O_hrightarrow D_{3d}rightarrow D_3$ is numerically observed as the length scale is increased: At short and intermediate length scales, the system behaves as having a rank-2 spin nematic order with $O_hrightarrow D_{3d}$ symmetry breaking; and at long distances, time reversal symmetry is further broken leading to the $O_hrightarrow D_3$ symmetry breaking. Finally, there is no numerical signature of spin orderings nor Luttinger liquid behaviors in the Kitaev phase whose nature is worth further studies.
Online recommendation and advertising are two major income channels for online recommendation platforms (e.g. e-commerce and news feed site). However, most platforms optimize recommending and advertising strategies by different teams separately via d
ifferent techniques, which may lead to suboptimal overall performances. To this end, in this paper, we propose a novel two-level reinforcement learning framework to jointly optimize the recommending and advertising strategies, where the first level generates a list of recommendations to optimize user experience in the long run; then the second level inserts ads into the recommendation list that can balance the immediate advertising revenue from advertisers and the negative influence of ads on long-term user experience. To be specific, the first level tackles high combinatorial action space problem that selects a subset items from the large item space; while the second level determines three internally related tasks, i.e., (i) whether to insert an ad, and if yes, (ii) the optimal ad and (iii) the optimal location to insert. The experimental results based on real-world data demonstrate the effectiveness of the proposed framework. We have released the implementation code to ease reproductivity.
We study the phase diagram of a one-dimensional version of the Kitaev spin-1/2 model with an extra ``$Gamma$-term, using analytical, density matrix renormalization group and exact diagonalization methods. Two intriguing phases are found. In the gaple
ss phase, although the exact symmetry group of the system is discrete, the low energy theory is described by an emergent SU(2)$_1$ Wess-Zumino-Witten (WZW) model. On the other hand, the spin-spin correlation functions exhibit SU(2) breaking prefactors, even though the exponents and the logarithmic corrections are consistent with the SU(2)$_1$ predictions. A modified nonabelian bosonization formula is proposed to capture such exotic emergent ``partial SU(2) symmetry. In the ordered phase, there is numerical evidence for an $O_hrightarrow D_4$ spontaneous symmetry breaking.
Video-based person re-identification (ReID) is a challenging problem, where some video tracks of people across non-overlapping cameras are available for matching. Feature aggregation from a video track is a key step for video-based person ReID. Many
existing methods tackle this problem by average/maximum temporal pooling or RNNs with attention. However, these methods cannot deal with temporal dependency and spatial misalignment problems at the same time. We are inspired by video action recognition that involves the identification of different actions from video tracks. Firstly, we use 3D convolutions on video volume, instead of using 2D convolutions across frames, to extract spatial and temporal features simultaneously. Secondly, we use a non-local block to tackle the misalignment problem and capture spatial-temporal long-range dependencies. As a result, the network can learn useful spatial-temporal information as a weighted sum of the features in all space and temporal positions in the input feature map. Experimental results on three datasets show that our framework outperforms state-of-the-art approaches by a large margin on multiple metrics.
