In computational advertising, platforms are increasingly optimizing toward advertisers' real assessment metrics to help achieve more reliable advertising performance. Consequently, predicting customers' Lifetime Value (LTV) has become an essential component of the advertising system, as it directly impacts the actual Return On Investment (ROI) of advertisers. Recent research on LTV prediction primarily focuses on the ranking stage, lacking consideration of the initial retrieval stage. This oversight may lead to the inconsistency between retrieval and ranking, resulting in a loss of efficiency. Unlike the LTV estimation in the ranking stage, the retrieval stage faces more severe data sparsity and constraints inherent in online scoring. Incorporating rich data from other domains can mitigate the sparsity while introducing the negative transfer issue. To tackle these challenges, we introduce Retrieval-LTV, a two-tower retrieval model for LTV prediction. This model employs a cooperative framework and incorporates a fine-grained evaluation for each sample across each expert, thereby enhancing effective selective learning from the source domain while mitigating the risk of negative transfer. Additionally, we have designed a specialized representation transformation to obtain the LTV-oriented score for online retrieval. Experiments on three real-world industrial datasets demonstrate that Retrieval-LTV outperforms all the baselines, achieving superior performance. An online A/B test further confirms the effectiveness of Retrieval-LTV, increasing the overall LTV by 2.08%. As a result, Retrieval-LTV has now been fully deployed in Tencent Ads.

The widespread adoption of location-based applications has created a growing demand for point-of-interest (POI) recommendation, which aims to predict a user’s next POI based on their historical check-in data and current location. However, existing methods often struggle to capture the intricate relationships within check-in data. This is largely due to their limitations in representing temporal and spatial information and underutilizing rich semantic features. While large language models (LLMs) offer powerful semantic comprehension to solve them, they are limited by hallucination and the inability to incorporate global collaborative information. To address these issues, we propose a novel method SeCor, which treats POI recommendation as a multi-modal task and integrates semantic and collaborative representations to form an efficient hybrid encoding. SeCor first employs a basic collaborative filtering model to mine interaction features. These embeddings, as one modal information, are fed into LLM to align with semantic representation, leading to efficient hybrid embeddings. To mitigate the hallucination, SeCor recommends based on the hybrid embeddings rather than directly using the LLM’s output text. Extensive experiments on three public real-world datasets show that SeCor outperforms all baselines, achieving improved recommendation performance by effectively integrating collaborative and semantic information through LLMs.

Seismic data interpolation is a vital technology for improving seismic data density. In recent years, deep learning approaches have demonstrated significant potential in this field, yielding impressive results. Nonetheless, challenges still persist and have not been adequately addressed. First, the lack of reliable labeled training datasets induces concerns about the network’s adaptiveness under supervised learning schemes. Additionally, due to inadequate spatial sampling, aliasing frequently poses considerable difficulties for deep neural networks. In this study, we tackle the issue of aliased seismic data interpolation through self-supervised learning. A novel dip-informed neural network (DINN) is introduced to explicitly integrate local dip information into the neural network and regularize the reconstruction of missing traces. To address the training challenges associated with regularly sampled seismic data interpolation under self-supervised learning schemes, a randomized mix training algorithm is developed. The experimental results along with comparisons to existing methods using both synthetic and field datasets demonstrate the effectiveness and robustness of our approach.

Accurately localizing 1D signal patterns, such as Gamma-ray well-log depth matching, is crucial in the oilfield service industry as it directly affects the quality of oil and gas exploration. However, traditional methods such as well-log curve analysis and pattern hand-picking matching are labor-intensive and heavily rely on human expertise, leading to inconsistent results. Although attempts have been made to automate this process, challenges such as low computational performance, non-robustness, and non-generalization remain unsolved. To address these challenges, we have developed a data-driven AI system that learns an active signal pattern localization strategy inspired by human attention. Our artificial intelligence system uses an offline reinforcement learning (RL) framework as its central component, which solves a highly abstracted Markov decision process problem via offline training on human-labeled historical data. The RL agent uses top-down reasoning to determine the location of target signal fragments by deforming a bounding window using simple transformation actions. To overcome distribution shifts between logged data and real and ensure generalization, we propose a discrete distributionally robust soft actor-critic RL framework (DRSAC-Discrete) to solve the Markov decision process problem under uncertainty. By exploring unfamiliar environments in a restrictive manner, the DRSAC-Discrete algorithm provides a safe solution that can be used when data is limited during the early stage of this industrial application. We evaluated the reinforcement learning-based localization system on augmented field Gamma-ray well-log datasets, and the results showed promising localization capability. Furthermore, the DRSAC-Discrete algorithm demonstrated relatively robust performance guarantees when facing data shortage.

The simultaneous-source technology for high-density seismic acquisition is a key solution to efficient seismic surveying. It is a cost-effective method when blended subsurface responses are recorded within a short time interval using multiple seismic sources. A following deblending process, however, is needed to separate signals contributed by individual sources. Recent advances in deep learning and its data-driven approach toward feature engineering have led to many new applications for a variety of seismic processing problems. It is still a challenge, though, to collect enough labeled data and avoid model overfitting and poor generalization performance over different datasets with a low resemblance from each other. In this article, we propose a novel self-supervised learning method to solve the deblending problem without labeled training datasets. Using a blind-trace deep neural network and a carefully crafted blending loss function, we demonstrate that the individual source-response pairs can be accurately separated under three different blended-acquisition designs.

Obtaining the latest Threat Intelligence (TI) via Twitter has become one of the most important methods for defenders to catch up with emerging cyber threats. Existing TI Twitter classification works mainly based on supervised learning methods. Such approaches require large amounts of annotated data and are difficult to be transferred to other TI Twitter classification tasks. This paper proposes a prompt-based method for classifying TI on Twitter, named TI-Prompt. TI-Prompt lever-ages the prompt-tuning method with two templates in different TI Twitter classification tasks. TI-Prompt also uses a semantic similarity-based approach to automatically enrich the prompt verbalizer without expert knowledge and a verbalizer refinement method to calibrate the verbalizer based on the training data. We evaluate TI-Prompt with binary and multi-classification tasks on two Twitter Threat Intelligence datasets. Evaluation results show that the proposed TI-Prompt improves 5-10% over the best performance of previous supervised learning methods under the few-shot settings. Compared to the general prompt-tuning methods, the proposed prompt-tuning templates can also improve the classification performance by 2–5%. Meanwhile, the proposed verbalizer enrichment method and refinement method improve classification accuracy by 1–4% compared with the general single-word verbalizer prompt method. Therefore, TI-Prompt can be extended to other Threat Intelligence classification tasks without requiring large amounts of training data, significantly reducing the annotation cost.

Higher dimensional (i.e., 2-D and 3-D) modeling is indispensable to correctly evaluate the responses of electromagnetic (EM) logging tools in complex formation environments. However, limited by the high computational cost of rigorous modeling, such as the finite-difference method and the finite-element method, the real-time applications in the well logging industry primarily rely on the 1-D forward solver, which would result in erroneous formation evaluation for complex scenarios. As a result, aiming at realizing fast modeling for EM logging tools in complex formations, this letter proposes a general framework assisted by deep neural networks (DNNs). The framework consists of three modules: earth model classification, parameter extraction, and surrogate construction. Separate DNNs are trained and tested for different modules. The accuracy and efficiency of the DNN-assisted fast modeling are validated by several experiments. This study finds that the fast modeling assisted by DNNs is able to calculate the tool responses and reconstruct the subsurface formations in real time.

The low-frequency seismic data provide crucial information for guiding the full-waveform inversion (FWI), especially when strong reflectors exist in the velocity model. However, hardware limitations make it difficult to acquire low-frequency data. To overcome the nonlinearity and ill-posedness caused by the absence of the low-frequency data, we develop an efficient progressive transfer learning algorithm for low-frequency extrapolation. The proposed method combines the FWI, the sparsity-promoted bandwidth-extension (BWE) algorithm, and the physics-guided data-driven deep learning approach. Compared with pure data-driven learning-based methods and the original progressive transfer learning method without BWE, our proposed algorithm shows better generalization ability. By integrating the physics constraints and the BWE algorithm, the performance of our method is less dependent on the quality of the initial training velocity model and the corresponding training set. We propose a logarithmic transformation to rebalance the loss function to overcome the challenge of predicting the weak reflection low-frequency data. To accelerate the algorithm, we propose a learning-based BWE method for initializing the training set and a truncated FWI method to reduce the iterative workflow’s computational cost. Experimental results show that our method achieves both high efficiency and high accuracy. The subsurface structures below the strong reflectors are successfully reconstructed with our extrapolated low-frequency data.

Reconstruction of seismic data is an important but challenging task in seismic data processing. Different machine-learning-based algorithms have been developed to solve this ill-posed problem and achieved great progress. However, most machine-learning-based methods rely on supervised learning where a good training dataset with many complete shot-gathers are required to train the model. Although the generative model has been used for unsupervised learning and reconstructing signals in a shot-gather, it fails to accurately resolve the fine features, especially when aliasing is the main concern. In addition, multiple shots’ interpolation problems have not been fully investigated by the unsupervised machine-learning-based approaches. In this work, we propose a self-supervised learning method using a blind-trace network and two antialiasing techniques (automatic spectrum suppression and mix-training) for seismic data reconstruction. The method is validated using challenging and realistic scenarios. Test results show that the method can be applied to single-shot or multiple shots’ cases and adapt well to different decimation patterns.