Robust image retrieval by cascading a deep quality assessment network

The performance of computer vision algorithms can severely degrade in the presence of a variety of distortions. While image enhancement algorithms have evolved to optimize image quality as measured according to human visual perception, their relevance in maximizing the success of computer vision algorithms operating on the enhanced image has been much less investigated. We consider the problem of image enhancement to combat Gaussian noise and low resolution with respect to the specific application of image retrieval from a dataset. We define the notion of image quality as determined by the success of image retrieval and design a deep convolutional neural network (CNN) to predict this quality. This network is then cascaded with a deep CNN designed for image denoising or super resolution, allowing for optimization of the enhancement CNN to maximize retrieval performance. This framework allows us to couple enhancement to the retrieval problem. We also consider the problem of adapting image features for robust retrieval performance in the presence of distortions. We show through experiments on distorted images of the Oxford and Paris buildings datasets that our algorithms yield improved mean average precision when compared to using enhancement methods that are oblivious to the task of image retrieval. ¹     

一种融合噪声网络的深度强化学习通信干扰资源分配算法

针对传统干扰资源分配算法在处理非线性组合优化问题时需要较完备的先验信息,同时决策维度小,无法满足现代通信对抗要求的问题,该文提出一种融合噪声网络的深度强化学习通信干扰资源分配算法(FNNDRL)。借鉴噪声网络的思想,该算法设计了孪生噪声评估网络,在避免Q值高估的基础上,通过提升评估网络的随机性,保证了训练过程的探索性;基于概率熵的物理意义,设计了基于策略分布熵改进的策略网络损失函数,在最大化累计奖励的同时最大化策略分布熵,避免策略优化过程中收敛到局部最优。仿真结果表明,该算法在解决干扰资源分配问题时优于所对比的平均分配和强化学习方法,同时算法稳定性较高,对高维决策空间适应性强。     

基于三维凹包学习算法的海面小目标检测方法

对于特征类的海面小目标检测方法,现有3特征检测器通常采用凸包分类算法完成检测。在实际应用时发现,该分类算法生成的判决区域在某些情况下不能很好地反映海杂波样本集合在特征空间中的分布情况,进而对检测器性能造成一定程度的损失。相比之下,使用凹包算法生成的判决区域是由凸包内剖得到的,它能更加贴合海杂波样本的分布,因此该文将判决区域的形式由凸包转化为凹包,并在此基础之上提出一种基于3维凹包学习算法的海面小目标检测方法。同时,针对现有3维凹包算法存在的内剖效率低、无法实现恒虚警检测的缺点,该文通过优化内剖点选择方法、增加“外补”环节的方式对算法进行改进。最后,经实测CSIR数据及X波段试验雷达数据共同验证,在其他参数均相同时,该文方法的检测性能要优于已有的多特征检测方法,并且通过对凹包算法的复杂度分析证明了所提方法的应用潜力。     

Deep ensemble-based hard sample mining for food recognition

Deep neural networks represent a compelling technique to tackle complex real-world problems, but are over-parameterized and often suffer from over- or under-confident estimates. Deep ensembles have shown better parameter estimations and often provide reliable uncertainty estimates that contribute to the robustness of the results. In this work, we propose a new metric to identify samples that are hard to classify. Our metric is defined as coincidence score for deep ensembles which measures the agreement of its individual models. The main hypothesis we rely on is that deep learning algorithms learn the low-loss samples better compared to large-loss samples. In order to compensate for this, we use controlled over-sampling on the identified ”hard” samples using proper data augmentation schemes to enable the models to learn those samples better. We validate the proposed metric using two public food datasets on different backbone architectures and show the improvements compared to the conventional deep neural network training using different performance metrics.     

关于海面无线传播模型的探讨

现有文献中一般都认为海面环境的无线传播模型近似于自由空间传播模型，但在实际测试中发现，自由空间模型预测值与实测值差别较大，不能很好地解释此类环境的信号传播。笔者根据多次海面覆盖测试的数据，分析并拟合出了新的海面传播模型。     

���ҳ����ˮƽ�����ѹ�Ѽ���

??Deep shale gas reservoirs buried underground with depth being more than 3 500 m are characterized by high in-situ stress, large horizontal stress difference, complex distribution of bedding and natural cracks, and strong rock plasticity. Thus, during hydraulic fracturing, these reservoirs often reveal difficult fracture extension, low fracture complexity, low stimulated reservoir volume (SRV), low conductivity and fast decline, which hinder greatly the economic and effective development of deep shale gas. In this paper, a specific and feasible technique of volume fracturing of deep shale gas horizontal wells is presented. In addition to planar perforation, multi-scale fracturing, full-scale fracture filling, and control over extension of high-angle natural fractures, some supporting techniques are proposed, including multi-stage alternate injection (of acid fluid, slick water and gel) and the mixed- and small-grained proppant to be injected with variable viscosity and displacement. These techniques help to increase the effective stimulated reservoir volume (ESRV) for deep gas production. Some of the techniques have been successfully used in the fracturing of deep shale gas horizontal wells in Yongchuan, Weiyuan and southern Jiaoshiba blocks in the Sichuan Basin. As a result, Wells YY1HF and WY1HF yielded initially 14.1×104 m3/d and 17.5×104 m3/d after fracturing. The volume fracturing of deep shale gas horizontal well is meaningful in achieving the productivity of 50×108 m3 gas from the interval of 3 500–4 000 m in Phase II development of Fuling and also in commercial production of huge shale gas resources at a vertical depth of less than 6 000 m.     

GEOCHEMICAL CHARACTERISTICS OF THE SKRUGARD OIL DISCOVERY,BARENTS SEA,ARCTIC NORWAY: A “PALAEO‐BIODEGRADED – GAS REACTIVATED” HYDROCARBON ACCUMULATION

This paper investigates the filling history of the Skrugard and Havis structures of the Johan Castberg field in the Polheim Sub‐Platform and Bjørnøyrenna Fault Complex, Barents Sea (Arctic Norway). Oil and gas occurs in the Early Jurassic and Middle Jurassic Nordmela and Stø Formations at Johan Castberg, and both free oil and bitumen are interpreted to be sourced from the Upper Jurassic Hekkingen Formation (Kimmeridge Formation equivalent). The geochemical characteristics of the petroleum from Skrugard and Havis, including the GOR, API and facies and maturity signatures, can be understood within a complex fill history which includes a palaeo oil charge, Tertiary uplift (>2 km), dismigration, in‐reservoir biodegradation, and late‐stage refill with gas. The API and GOR of the Skrugard oil are 31° and 60m³/m³, respectively. The petroleum is geochemically similar to that in the nearby Havis structure, to that in the Snøhvit region to the south of the Loppa High, and also to the petroleum recorded as traces in well 7219/9‐1, approximately 16 km SW of Johan Castberg field. However, the petroleum differs from the oil in the Alta well 7120/2‐1, located in the southern part of the Loppa High, illustrating the complexity of the regional petroleum systems. The Skrugard oil is of medium maturity (ca. 0.8–0.9% R_c), and is significantly biodegraded despite being gas‐saturated. Evidence for biodegradation includes the reduced concentrations of C₁₀‐C₂₅ n‐alkanes and the presence of a prominent unresolved complex mixture (UCM) in gas chromatogram traces. However non‐biodegraded C₄‐C₈ range hydrocarbons are also present in the reservoir. This suggests a recent charge of gas/condensate into the structure which therefore contains a mixture of palaeo‐degraded and unaltered petroleum. Oil‐type inclusions within authigenic quartz and feldspar from reservoir sandstones at Skrugard were analysed. The results indicate that the structure (present‐day depth 1276–1395m) underwent Tertiary uplift by ca. 2–3km following an earlier phase of oil emplacement. The presence of the oil type inclusions, both in the current gas zone (Stø Formation) and in the oil zone (Stø and Nordmela Formations), indicates that the positions of the oil‐water and gas‐oil contacts have changed over time. This is consistent with a recent gas charge to the upper part of the reservoir, and also with the gas being at dew point. These observations are supported by analyses of core extracts which show an increasing bitumen content towards the OWC, and the oil‐type bitumen in the present‐day gas zone. A charge history model for the Skrugard structure is proposed which integrates both the observations concerning the petroleum inclusions and the biodegraded oil together with observations of seismically‐monitored gas fluxes along the rim of the Loppa High. Improved understanding of the Skrugard structure and its filling history will assist exploration in similar settings in other parts of the Barents Sea and worldwide, particularly where multiple source rocks and a multi‐stage charge history have controlled reservoir filling.     

Trapping and recombination processes via deep level T3 in semi-insulating gallium arsenide

Trapping and recombination of free carriers by deep level T₃ has been studied. Occupancy of the level by electrons and dynamics of its filling and emptying as a function of illumination with monoenergetic photons in 0.69–1.55 eV range has been monitored by the thermally stimulated currents method. We have found that level T₃ behaves more like a recombination center than like an ordinary electron trap. Besides trapping free electrons from conduction band, this trap can also communicate with valence band, trapping holes. The capture cross section for trapping a hole is estimated to be comparable or even larger than the capture cross section for trapping an electron. However, in many experimental conditions free electrons are generated more abundantly than free holes, and free carrier mobility and thermal velocity are both much higher for electrons than for holes. Therefore, electron trapping often prevails, so that this frequently detected defect, has been up to now most often perceived as a deep electron trap.     

深亚微米薄膜SOI MOSFET的二维数值模拟

本文介绍了适合于薄膜亚微米、深亚微米ＳＯＩＭＯＳＦＥＴ的二维数值模拟软件。该模拟软件同时考虑了两种载流子的产生－复合作用，采用了独特的动态二步法求解泊松方程和电子、空穴的电流连续性方程，提高了计算效率和收敛性。利用此模拟软件较为详细地分析了薄膜ＳＯＩＭＯＳＦＥＴ不同于厚膜ＳＯＩＭＯＳＦＥＴ的工作机理及特性，发现薄膜ＳＯＩＭＯＳＦＥＴ的所有特性几乎都得到了改善。将模拟结果与实验结果进行了对比，两者吻合得较好。     

MOS结构深能级瞬态谱的测试与分析

叙述了一种采用MOS结构研究半导体深能级陷阱的深能级瞬态谱(DLTS)测试与分析方法。该方法简便易行,适用范围广。通过对n-Al_(0.2)Ga_(0.8)As中深能级的研究表明,MOS结构和p~+-n结构的DLTS结果相同。