Forward-looking sonar (FLS) compresses elevation information in 2-D imaging. This makes the recovery of elevation-related structure from a single FLS image an underdetermined problem. Existing pseudo front view (PFV)-based single-frame methods have shown potential, but most still rely on generic isotropic 2-D encoding so known range-azimuth directional characteristics are not sufficiently integrated into PFV recovery. To address this limitation, this letter proposes a direction-aware structured inversion method for single-frame FLS elevation recovery. The method enhances PFV recovery through direction-decoupled feature extraction, azimuth-preserving encoding, and target-aware directional modulation. This improves the representation of geometry-related structural cues in elevation-compressed acoustic observations. Experiments on two public datasets show that the proposed method outperforms existing PFV-based methods and produces more accurate PFV predictions and 3-D acoustic inversion results. Qualitative results on real deep-sea FLS data further support its practical applicability by showing response-consistent 3-D acoustic inversion in measured sonar scenes.

Marine oil spills pose significant threats to ecosystems and human health, emphasizing the importance of synthetic aperture radar (SAR) images for reliable and all-weather monitoring. However, current methods face two major challenges. The first is data limitations, including insufficient data quantity and noise, such as speckle noise and distortions introduced during preprocessing. The second is look-alike confusions, which pose challenges in distinguishing oil spills from visually similar phenomena. This article introduces TransOilSeg, a novel method designed to address these challenges and enhance oil spill detection performance. TransOilSeg employs a transfer learning component (TLC) to integrate data from diverse geographical regions and varying quality, learning general features from multisource datasets. By leveraging a gradient aggregation algorithm, the model combines features from limited and noisy SAR oil spill (SOS) datasets, transferring data deficiencies. In addition, the adaptive attention hybrid encoder (AAHE) analyzes contextual features and adapts to varying datasets, enabling the model to effectively distinguish oil spills from look-alike phenomena. Comprehensive evaluations across multiple datasets demonstrate the robust generalization capability of TransOilSeg. On the M4D dataset, which includes 1002 training samples, the model achieved a mean intersection over union (mIoU) of 61.38% for oil spill detection and 62.41% for look-alike detection. Furthermore, TransOilSeg maintained strong performance when transferred between datasets with varying levels of noise and distortions, demonstrating its adaptability to challenging conditions. These results highlight its potential as a reliable tool for marine oil spill detection and monitoring.

Automated software debugging plays a crucial role in aiding software developers to swiftly identify and attempt to rectify faults, thereby significantly reducing developers’ workload. Previous researches have predominantly relied on simplistic semantic deep learning or statistical analysis methods to locate faulty statements in diverse projects. However, code repositories often consist of lengthy sequences with long-distance dependencies, posing challenges for accurately modeling fault localization using these methods. In addition, the lack of joint reasoning among various faults prevents existing models from deeply capturing fault information. To address these challenges, we propose a method named CodeHealer to achieve accurate fault localization and program repair. CodeHealer comprises three components: a Deep Semantic Information Extraction Component that effectively extracts deep semantic features from suspicious code statements using classifiers based on Joint-attention mechanisms; a Suspicious Statement Ranking Component that combines various fault localization features and employs multilayer perceptrons to derive multidimensional vectors of suspicion values; and a Fault Repair Component that, based on ranked suspicious statements generated by fault localization, adopts a top-down approach using multiple classifiers based on Co-teaching mechanisms to select repair templates and generate patches. The experimental results indicate that when applied to fault localization, CodeHealer outperforms the best baseline method with improvements of 11.4%, 2.7%, and 1.6% on Top-1/3/5 metrics, respectively. It also reduces the MFR and MAR by 9.8% and 2.1%, where lower values denote better fault localization effectiveness. Additionally, in automated software debugging, CodeHealer fixes an additional 6 faults compared to the current best method, totaling 53 faults repaired.

China, now the global leader in energy production and consumption, has made significant strides towards its dual-carbon goals as of 2021. This underlines the crucial need for accurate regional carbon emission forecasting. Traditional back propagation neural networks face challenges with local minima and slow convergence. To overcome these, our study proposes an innovative method that combines LSTM-Attention, enriched with self-attention mechanisms, and Elman neural networks with the traditional BP neural model for regional carbon emission prediction. Through the development of a predictive model leveraging LSTM-Attention and Elman neural networks, and applying it to regional carbon emission forecasts, our findings reveal that this model achieves higher accuracy than conventional BP neural networks and substantially reduces the need for gas sensors, thus lowering costs in the statistical process.

Abstract Source code summaries improve the readability and intelligibility of code, help developers understand programs, and improve the efficiency of software maintenance and upgrade processes. Unfortunately, these code comments are often mismatched, missing, or outdated in software projects, resulting in developers needing to infer functionality from source code, affecting the efficiency of software maintenance and evolution. Various methods based on neuronal networks are proposed to solve the problem of synthesis of source code. However, the current work is being carried out on resource‐rich programming languages such as Java and Python, and some low‐resource languages may not perform well. In order to solve the above challenges, we propose a context‐based transfer learning model for low resource code summarization (LRCS), which learns the common information from the language with rich resources, and then transfers it to the target language model for further learning. It consists of two components: the summary generation component is used to learn the syntactic and semantic information of the code, and the learning transfer component is used to improve the generalization ability of the model in the learning process of cross‐language code summarization. Experimental results show that LRCS outperforms baseline methods in code summarization in terms of sentence‐level BLEU, corpus‐level BLEU and METEOR. For example, LRCS improves corpus‐level BLEU scores by 52.90%, 41.10%, and 14.97%, respectively, compared to baseline methods.