Oil wells play an important role in the extraction of oil and gas, and their future potential extends beyond oil and gas exploitation to include the development of geothermal resources for sustainable power generation. Identifying and detecting oil wells are of paramount importance given the crucial role of oil well distribution in energy planning. In recent years, significant progress has been made in detecting single oil well objects, with recognition accuracy exceeding 90%. However, there are still remaining challenges, particularly with regard to small-scale objects, varying viewing angles, and complex occlusions within the domain of oil well detection. In this work, we created our own dataset, which included 722 images containing 3749 oil well objects in Daqing, Huatugou, Changqing oil field areas in China, and California in the USA. Within this dataset, 2165 objects were unoccluded, 617 were moderately occluded, and 967 objects were severely occluded. To address the challenges in detecting oil wells in complex occlusion scenarios, we propose the YOLOv5s-seg CAM NWD network for object detection and instance segmentation. The experimental results show that our proposed model outperforms YOLOv5 with F1 improvements of 5.4%, 11.6%, and 23.1% observed for unoccluded, moderately occluded, and severely occluded scenarios, respectively.

Background: Liver cancer (Hepatocellular carcinoma; HCC) prevalence is increasing and with poor clinical outcome expected it means greater understanding of HCC aetiology is urgently required. This study explored a deep learning solution to detect biologically important features that distinguish prognostic subgroups. A novel architecture of an Artificial Neural Network (ANN) trained with a customised objective function (L_RSC) was developed. The ANN should discover new data representations, to detect patient subgroups that are biologically homogenous (clustering loss) and similar in survival (survival loss) while removing noise from the data (reconstruction loss). The model was applied to TCGA-HCC multi-omics data and benchmarked against baseline models that only use a reconstruction objective function (BCE, MSE) for learning. With the baseline models, the new features are then filtered based on survival information and used for clustering patients. Different variants of the customised objective function, incorporating only reconstruction and clustering losses (L_RC); and reconstruction and survival losses (L_RS) were also evaluated. Robust features consistently detected were compared between models and validated in TCGA and LIRI-JP HCC cohorts. Results: The combined loss (L_RSC) discovered highly significant prognostic subgroups (P-value = 1.55E−77) with more accurate sample assignment (Silhouette scores: 0.59–0.7) compared to baseline models (0.18–0.3). All L_RSC bottleneck features (N = 100) were significant for survival, compared to only 11–21 for baseline models. Prognostic subgroups were not explained by disease grade or risk factors. Instead L_RSC identified robust features including 377 mRNAs, many of which were novel (61.27%) compared to those identified by the other losses. Some 75 mRNAs were prognostic in TCGA, while 29 were prognostic in LIRI-JP also. L_RSC also identified 15 robust miRNAs including two novel (hsa-let-7g; hsa-mir-550a-1) and 328 methylation features with 71% being prognostic. Gene-enrichment and Functional Annotation Analysis identified seven pathways differentiating prognostic clusters. Conclusions: Combining cluster and survival metrics with the reconstruction objective function facilitated superior prognostic subgroup identification. The hybrid model identified more homogeneous clusters that consequently were more biologically meaningful. The novel and prognostic robust features extracted provide additional information to improve our understanding of a complex disease to help reveal its aetiology. Moreover, the gene features identified may have clinical applications as therapeutic targets.

Record linkage, referred to also as entity resolution, is the process of identifying pairs of records representing the same real world entity (e.g. a person) within a dataset or across multiple datasets. In order to reduce the number of record comparisons, record linkage frameworks initially perform a process referred to as blocking, which involves splitting records into a set of blocks using a partition (or blocking) scheme. This restricts comparisons among records that belong to the same block during the linkage process. Existing blocking methods are often evaluated using different metrics and independently of the choice of the subsequent linkage method, which makes the choice of an optimal approach very subjective. In this paper we demonstrate that existing evaluation metrics fail to provide strong evidence to support the selection of an optimal blocking method. We conduct an extensive evaluation of different blocking methods using multiple datasets and some commonly applied linkage techniques to show that evaluation of a blocking method must take into consideration the subsequent linkage phase. We propose a novel evaluation technique that takes into consideration multiple factors including the end-to-end running time of the combined blocking and linkage phases as well as the linkage technique used. We empirically demonstrate using multiple datasets that according to this novel evaluation technique some blocking methods can be fairly considered superior to others, while some should be deemed incomparable according to those factors. Finally, we propose a novel blocking method selection procedure that takes into consideration the linkage proficiency and end-to-end time of different blocking methods combined with a given linkage technique. We show that this technique is able to select the best or near best blocking method for unseen data.

There have been massive advances in AI technologies towards addressing the contemporary challenge of fake news identification. However, these technologies, as observed widely, have not had the same kind or depth in impact across global societies. In particular, the AI scholarship in fake news detection arguably has not been as beneficial or appropriate for Global South, bringing geo-political bias into the picture. While it is often natural to think of data bias as the potential reason for geo-political bias, other factors could be much more important in being more latent, and thus less visible. In this commentary, we investigate as to how the facet of affect, comprising emotions and sentiments, could be a potent vehicle for geo-political biases in AI. We highlight, through assembling and interpreting insights from literature, the overarching neglect of affect across methods for fake news detection AI, and how this could be a potentially important factor for geo-political bias within them. This exposition, we believe, also serves as a first effort in understanding how geo-political biases work within AI pipelines beyond the data collection stage.