FlipDA: Effective and Robust Data Augmentation for Few-Shot Learning

Most previous methods for text data augmentation are limited to simple tasks and weak baselines. We explore data augmentation on hard tasks (i.e., few-shot natural language understanding) and strong baselines (i.e., pretrained models with over one billion parameters). Under this setting, we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much. To address this challenge, we propose a novel data augmentation method FlipDA that jointly uses a generative model and a classifier to generate label-flipped data. Central to the idea of FlipDA is the discovery that generating label-flipped data is more crucial to the performance than generating label-preserved data. Experiments show that FlipDA achieves a good tradeoff between effectiveness and robustness---it substantially improves many tasks while not negatively affecting the others.

Photoacoustic Silk Scaffolds for Neural stimulation and Regeneration

Neural interfaces using biocompatible scaffolds provide crucial properties for the functional repair of nerve injuries and neurodegenerative diseases, including cell adhesion, structural support, and mass transport. Neural stimulation has also been found to be effective in promoting neural regeneration. This work provides a new strategy to integrate photoacoustic (PA) neural stimulation into hydrogel scaffolds using a nanocomposite hydrogel approach. Specifically, polyethylene glycol (PEG)-functionalized carbon nanotubes (CNT), highly efficient photoacoustic agents, are embedded into silk fibroin to form biocompatible and soft photoacoustic materials. We show that these photoacoustic functional scaffolds enable non-genetic activation of neurons with a spatial precision defined by the area of light illumination, promoting neuron regeneration. These CNT/silk scaffolds offered reliable and repeatable photoacoustic neural stimulation. 94% of photoacoustic stimulated neurons exhibit a fluorescence change larger than 10% in calcium imaging in the light illuminated area. The on-demand photoacoustic stimulation increased neurite outgrowth by 1.74-fold in a dorsal root ganglion model, when compared to the unstimulated group. We also confirmed that photoacoustic neural stimulation promoted neurite outgrowth by impacting the brain-derived neurotrophic factor (BDNF) pathway. As a multifunctional neural scaffold, CNT/silk scaffolds demonstrated non-genetic PA neural stimulation functions and promoted neurite outgrowth, providing a new method for non-pharmacological neural regeneration.

PHI-MVS: Plane Hypothesis Inference Multi-view Stereo for Large-Scale Scene Reconstruction

PatchMatch based Multi-view Stereo (MVS) algorithms have achieved great success in large-scale scene reconstruction tasks. However, reconstruction of texture-less planes often fails as similarity measurement methods may become ineffective on these regions. Thus, a new plane hypothesis inference strategy is proposed to handle the above issue. The procedure consists of two steps: First, multiple plane hypotheses are generated using filtered initial depth maps on regions that are not successfully recovered; Second, depth hypotheses are selected using Markov Random Field (MRF). The strategy can significantly improve the completeness of reconstruction results with only acceptable computing time increasing. Besides, a new acceleration scheme similar to dilated convolution can speed up the depth map estimating process with only a slight influence on the reconstruction. We integrated the above ideas into a new MVS pipeline, Plane Hypothesis Inference Multi-view Stereo (PHI-MVS). The result of PHI-MVS is validated on ETH3D public benchmarks, and it demonstrates competing performance against the state-of-the-art.

GPT Understands, Too

While GPTs with traditional fine-tuning fail to achieve strong results on natural language understanding (NLU), we show that GPTs can be better than or comparable to similar-sized BERTs on NLU tasks with a novel method P-tuning -- which employs trainable continuous prompt embeddings. On the knowledge probing (LAMA) benchmark, the best GPT recovers 64% (P@1) of world knowledge without any additional text provided during test time, which substantially improves the previous best by 20+ percentage points. On the SuperGlue benchmark, GPTs achieve comparable and sometimes better performance to similar-sized BERTs in supervised learning. Importantly, we find that P-tuning also improves BERTs performance in both few-shot and supervised settings while largely reducing the need for prompt engineering. Consequently, P-tuning outperforms the state-of-the-art approaches on the few-shot SuperGlue benchmark.

Learning Structral coherence Via Generative Adversarial Network for Single Image Super-Resolution

Among the major remaining challenges for single image super resolution (SISR) is the capacity to recover coherent images with global shapes and local details conforming to human vision system. Recent generative adversarial network (GAN) based SISR methods have yielded overall realistic SR images, however, there are always unpleasant textures accompanied with structural distortions in local regions. To target these issues, we introduce the gradient branch into the generator to preserve structural information by restoring high-resolution gradient maps in SR process. In addition, we utilize a U-net based discriminator to consider both the whole image and the detailed per-pixel authenticity, which could encourage the generator to maintain overall coherence of the reconstructed images. Moreover, we have studied objective functions and LPIPS perceptual loss is added to generate more realistic and natural details. Experimental results show that our proposed method outperforms state-of-the-art perceptual-driven SR methods in perception index (PI), and obtains more geometrically consistent and visually pleasing textures in natural image restoration.

Topological $R$-pressure and topological pressure of free semigroup actions

In this paper we introduce the definition of topological $r$-pressure of free semigroup actions on compact metric space and provide some properties of it. Through skew-product transformation into a medium, we can obtain the following two main results. 1. We extend the result that the topological pressure is the limit of topological $r$-pressure incite{C} to free semigroup actions ($rto 0$). 2. Let $f_i,$ $i=0, 1, cdots, m-1$, be homeomorphisms on a compact metric space. For any continuous function, we verify that the topological pressure of $f_0, cdots, f_{m-1}$ equals the topological pressure of $f_0^{-1}, cdots, f_{m-1}^{-1}.$

Wind Driven Semiconductor Electricity Generator With High Direct Current Output Based On a Dynamic Schottky Junction

As the fast development of internet of things (IoTs), distributed sensors have been frequently used and the small and portable power sources are highly demanded. However, the present portable power source such as lithium battery has low capacity and need to be replaced or recharged frequently. A portable power source which can continuously generate electrical power in situ will be an idea solution. Herein, we demonstrate a wind driven semiconductor electricity generator based on a dynamic Schottky junction, which can output a continuous direct current with an average value of 4.4 mA (the maximum value of 8.4 mA) over 360 seconds. Compared with the previous metal/semiconductor generator, the output current is one thousand times higher. Furthermore, this wind driven generator has been explored to function as a turn counter due to its stable output and also to drive a graphene ultraviolet photodetector, which shows a responsivity of 35.8 A/W under the 365 nm ultraviolet light. Our research provides a feasible method to achieve wind power generation and power supply for distributed sensors in the future.

Galaxy properties in the cosmic web of EAGLE simulation

We investigate the dependence of the galaxy properties on cosmic web environments using the most up-to-date hydrodynamic simulation: Evolution and Assembly of Galaxies and their Environments (EAGLE). The baryon fractions in haloes and the amplitudes of the galaxy luminosity function decrease going from knots to filaments to sheets to voids. Interestingly, the value of L$^*$ varies dramatically in different cosmic web environments. At z = 0, we find a characteristic halo mass of $10^{12} h^{-1}rm M_{odot}$, below which the stellar-to-halo mass ratio is higher in knots while above which it reverses. This particular halo mass corresponds to a characteristic stellar mass of $1.8times 10^{10} h^{-1}rm M_{odot}$. Below the characteristic stellar mass central galaxies have redder colors, lower sSFRs and higher metallicities in knots than those in filaments, sheets and voids, while above this characteristic stellar mass, the cosmic web environmental dependences either reverse or vanish. Such dependences can be attributed to the fact that the active galaxy fraction decreases along voids, sheets, filaments and knots. The cosmic web dependences get weaker towards higher redshifts for most of the explored galaxy properties and scaling relations, except for the gas metallicity vs. stellar mass relation.

Rises in forests of binary shrubs

The study of patterns in permutations associated with forests of binary shrubs was initiated by D. Bevan et al.. In this paper, we study five different types of rise statistics that can be associated with such permutations and find the generating functions for the distribution of such rise statistics.

Magnetohydrodynamic Simulation of the Interaction between Interplanetary Strong Shock and Magnetic Cloud and its Consequent Geoeffectiveness

Numerical studies have been performed to interpret the observed shock overtaking magnetic cloud (MC) event by a 2.5 dimensional magnetohydrodynamic (MHD) model in heliospheric meridional plane. Results of an individual MC simulation show that the MC travels with a constant bulk flow speed. The MC is injected with very strong inherent magnetic field over that in the ambient flow and expands rapidly in size initially. Consequently, the diameter of MC increases in an asymptotic speed while its angular width contracts gradually. Meanwhile, simulations of MC-shock interaction are also presented, in which both a typical MC and a strong fast shock emerge from the inner boundary and propagate along heliospheric equator, separated by an appropriate interval. The results show that the shock firstly catches up with the preceding MC, then penetrates through the MC, and finally merges with the MC-driven shock into a stronger compound shock. The morphologies of shock front in interplanetary space and MC body behave as a central concave and a smooth arc respectively. The compression and rotation of magnetic field serve as an efficient mechanism to cause a large geomagnetic storm. The MC is highly compressed by the the overtaking shock. Contrarily, the transport time of incidental shock influenced by the MC depends on the interval between their commencements. Maximum geoeffectiveness results from that when the shock enters the core of preceding MC, which is also substantiated to some extent by a corresponding simplified analytic model. Quantified by $Dst$ index, the specific result gives that the geoeffectiveness of an individual MC is largely enhanced with 80% increment in maximum by an incidental shock.