Existing stereo matching techniques often struggle with detailing subtle objects on depth edges. To alleviate this problem, we introduced the Dynamic-Range Disparity Initialization module, which integrates three complementary branches: the dynamic dense volume for localized disparity sampling, the sparse global volume for encoding search center information, and the background static volume with skip connections for enhancing depth edge accuracy. The dynamic dense volume identifies optimal search centers for each pixel and performs local neighborhood sampling for cost aggregation, thereby generating the initial disparity map. Since the search center positions get lost while assembling the disparity samples, a sparse global volume is proposed to implicitly encode these positions during the training process of the network. We also designed the Inception Update module based on our analysis of convolutional structures and nonlinear gating mechanisms. RetinaStereo achieves state-of-the-art performance on the KITTI-2015 leaderboard for the D1-fg metric among published methods.

Long-tail distribution includes two kinds of imbalance: the class-wise imbalance and the attribute-wise one. Sufficient studies have been conducted on the class-wise imbalance. However, the distribution with both the class-wise and the attribute-wise imbalance is more difficult since the attribute is unknown. To solve this problem, we introduce a novel Generalized Long-tail Exact Penalty Method (GL-EPM). GLEPM redefines the long-tail classification as a constrained optimization problem where the model’s prediction on all samples, both the head samples and the tail samples, are constrained to ally with their labels. We then tackles the constrained problem using the exact penalty method, transforming the problem into a series of unconstrained penalty problems. Each penalty problem in the series is generated by the solution of the previous one and the solution series can be theoretically proved to converge to the solution of the constrained optimization problem. We further solve the penalty problems with a rapid-update solution, where the model is trained with dynamically updated sampling rates and loss weights. To validate the efficacy of our proposed method, we conduct a fair comparison with existing baselines across all generalized long-tail benchmarks, demonstrating state-of-the-art performance.

Defocus deblurring is a classic problem in image restoration tasks. The formation of its defocus blur is related to depth. Recently, the use of dual-pixel sensor designed according to depth-disparity characteristics has brought great improvements to the defocus deblurring task. However, the difficulty of real-time acquisition of dual-pixel images brings difficulties to algorithm deployment. This inspires us to remove defocus blur by single image with depth information. We propose a single-image depth-enhanced defocus deblurring network, which uses a depth map estimated by the monocular depth estimation network to guide the network defocus deblurring. We design a deep information fusion unit, which greatly improves the effect of deblurring. Experiments show that on the single image defocus deblurring task, the experimental results demonstrate the superiority of our method.

Despite supervised deep stereo matching networks have achieved impressive performance given sufficient training data, the poor generalization ability caused by the domain shifts prevents them from being applied to unseen domains. Recent progress has shown that CLIP could be a promising alternative for zero-shot visual representation task under the natural language supervision. In this paper, we present a new framework for domain generalized stereo matching by leveraging the contrastive language-image pre-training (CLIP), which distills text-guided discriminative content information rather than task-irrelevant style information. Extensive experiments show that the model generalization ability can be improved significantly in the unseen domain when transferring from SceneFlow to Middlebury, ETH3D and KITTI.

Customer baseline load (CBL) reconstruction is a critical problem in residential demand response. The difficulty of residential CBL lies in the variability of both irregular consumption and on-site distributed energy resources. Targeting the CBL reconstruction of residential prosumers, a regression-based estimation scheme is proposed using stacked autoencoders (SAEs) under the federated learning (FL) framework. In the FL framework, each residential unit (RU) stores and trains data locally without sharing them with neighboring RUs or the independent third party (ITP) responsible for CBL reconstruction. Local updates containing no load information are exchanged with the ITP (server) for the training improvement. The FL framework can, thus, protect the privacy of customers. Experimental results show that the proposed FL-based cascaded SAE outperforms the baseline on all tests and achieves up to 62.5% improvement in reducing reconstruction error. Moreover, it has enhanced privacy-preserving knowledge-sharing ability, higher efficiency, and better stability.

Unexpected large power surges will cause instantaneous grid shock and, thus, emergency control plans must be implemented to prevent the system from collapsing. In this article, with the aid of reinforcement learning, novel model-free control (MFC)-based emergency control schemes are presented. First, multi-Q-learning-based emergency plans are designed for limited emergency scenarios by using offline-training-online-approximation methods. To solve the more general multiscenario emergency control problem, a deep deterministic policy gradient (DDPG) algorithm is adopted to learn near-optimal solutions. With the aid of deep Q network, DDPG-based strategies have better generalization abilities for unknown and untrained emergency scenarios, and thus are suitable for multiscenario learning. Through simulations using benchmark systems, the proposed schemes are proven to achieve satisfactory performances.

Battery swapping stations (BSSs) are ideal candidates for fast frequency regulation services (FFRS) due to their large battery stock capacity. In addition, BSSs can precharge batteries for customers and the batteries that are not in charging can provide a stable regulation capacity to the market. However, uncertainties, such as ACE signals and the EV per-hour visit counts, introduce stochastic nonlinear dynamics into the operation of a BSS-based FFRS. Currently, there is no quantification method to ensure its optimal economical operation. To close this gap, in this article, we propose a novel deep Q-learning-based FFRS capacity dynamic scheduling strategy. This method can autonomously schedule the hourly regulation capacity in real time to maximize the BSS's revenue for providing FFRS. Case studies using real-world data verify the efficacy of the proposed work.

Accurate estimation of customer baseline load (CBL) is a key factor in the successful implementation of demand response (DR). CBL technologies implemented at utilities currently are primarily designed for large industrial and commercial customers. The U.S. Federal Energy Regulatory Commission (FERC) order 745 states that DR owners, including residential customers, can sell their load reduction in the wholesale market. However, since residential load is random and un-schedulable, this tends to inherently degrade the effectiveness of existing CBL technologies. In this paper, a novel SAE based CBL method for residential customers that uses the data reconstruction capability of a stacked autoencoder (SAE) is described. In the model, two SAEs are synchronously trained-one SAE generates a pseudo-load pool and the second one is used to select a pseudo-load to reconstruct a residential CBL. A support vector machine (SVM) classifier is self-trained to conduct the pseudo-load selection. The proposed strategy is validated using a real data set consisting of 328 residential customers' smart meter readings. Benchmarks from other machine learning techniques and existing CBL methods are compared with the proposed method. Test results show that the accuracy of the residential CBL reconstruction significantly improves when compared with existing methods, such as HighXofY and exponential moving average.

Fast algorithms for nearest neighbor (NN) search have in large part focused on L2 distance. Here we develop an approach for L1 distance that begins with an explicit and exact embedding of the points into L2. We show how this embedding can efficiently be combined with random projection methods for L2 NN search, such as locality-sensitive hashing or random projection trees. We rigorously establish the correctness of the methodology and show by experimentation that it is competitive in practice with available alternatives.

We propose an efficient algorithmic framework for time-domain circuit simulation using exponential integrators. This work addresses several critical issues exposed by previous matrix exponential based circuit simulation research, and makes it capable of simulating stiff nonlinear circuit system at a large scale. In this framework, the system's nonlinearity is treated with exponential Rosenbrock-Euler formulation. The matrix exponential and vector product is computed using invert Krylov subspace method. Our proposed method has several distinguished advantages over conventional formulations (e.g., the well-known backward Euler with Newton-Raphson method). The matrix factorization is performed only for the conductance/resistance matrix G, without being performed for the combinations of the capacitance/inductance matrix C and matrix G, which are used in traditional implicit formulations. Furthermore, due to the explicit nature of our formulation, we do not need to repeat LU decompositions when adjusting the length of time steps for error controls. Our algorithm is better suited to solving tightly coupled post-layout circuits in the pursuit for full-chip simulation. Our experimental results validate the advantages of our framework.

Via-programmable gate arrays (VPGAs) offer a middle ground between application specific integrated circuits and field programmable gate arrays in terms of flexibility, manufactuing cost, speed, power and area. In this paper, we present a novel VPGA logic cell, the complementary universal logic gate (CULG) which can be used to implement both sequential and combinatorial elements. Its performance is compared with a number of other designs including transmission gate, differential cascode voltage switch with pass gate, and standard cell. The CULG is found to have comparable power-delay product and process variation sensitivity to the other designs while offering the lowest power consumption.