Video recommendation systems heavily rely on user watch time feedback, making accurate watch time prediction a crucial task. However, this task inherently suffers from bias, as recommendation models tend to favor long-duration videos to maximize watch time. This issue, known as duration bias in the watch-time prediction context, can be explained from a causal perspective, where video duration acts as a confounder. Recent works address this bias using backdoor adjustment, isolating the direct effect of content on watch time from observational data. These methods typically discretize video duration into groups, estimate group-wise effects, and then aggregate them via a unified prediction model. However, this aggregation strategy is prone to model misspecification due to feature distribution shift across groups. In this paper, we reinterpret the problem through the lens of invariant learning and propose a novel framework: Duration-Invariant Feature Learning (DIFL). DIFL employs a kernel-based regularization that enforces representation invariance across duration groups, reducing sensitivity to group design and improving generalization. This enables more accurate modeling of the direct causal effect and making counterfactual inference. Extensive experiments on both public and real large-scale production datasets demonstrate the effectiveness of our approach, which achieves SOTA performance.

Generative Recommenders (GRs), exemplified by the Hierarchical Sequential Transduction Unit (HSTU), have emerged as a powerful paradigm for modeling long user interaction sequences. However, we observe that their ''flat-sequence'' assumption overlooks the rich, intrinsic structure of user behavior. This leads to two key limitations: a failure to capture the temporal hierarchy of session-based engagement, and computational inefficiency, as dense attention introduces significant noise that obscures true preference signals within semantically sparse histories, which deteriorates the quality of the learned representations. To this end, we propose a novel framework named HPGR (Hierarchical and Preference-aware Generative Recommender), built upon a two-stage paradigm that injects these crucial structural priors into the model to handle the drawback. Specifically, HPGR comprises two synergistic stages. First, a structure-aware pre-training stage employs a session-based Masked Item Modeling (MIM) objective to learn a hierarchically-informed and semantically rich item representation space. Second, a preference-aware fine-tuning stage leverages these powerful representations to implement a Preference-Guided Sparse Attention mechanism, which dynamically constrains computation to only the most relevant historical items, enhancing both efficiency and signal-to-noise ratio. Empirical experiments on a large-scale proprietary industrial dataset from APPGallery and an online A/B test verify that HPGR achieves state-of-the-art performance over multiple strong baselines, including HSTU and MTGR.

Deep Click-Through Rate (CTR) prediction models play an important role in modern industrial recommendation scenarios. However, high memory overhead and computational costs limit their deployment in resource-constrained environments. Low-rank approximation is an effective method for computer vision and natural language processing models, but its application in compressing CTR prediction models has been less explored. Due to the limited memory and computing resources, compression of CTR prediction models often confronts three fundamental challenges, i.e., (1). How to reduce the model sizes to adapt to edge devices? (2). How to speed up CTR prediction model inference? (3). How to retain the capabilities of original models after compression? Previous low-rank compression research mostly uses tensor decomposition, which can achieve a high parameter compression ratio, but brings in AUC degradation and additional computing overhead. To address these challenges, we propose a unified low-rank decomposition framework for compressing CTR prediction models. We find that even with the most classic matrix decomposition SVD method, our framework can achieve better performance than the original model. To further improve the effectiveness of our framework, we locally compress the output features instead of compressing the model weights. Our unified low-rank compression framework can be applied to embedding tables and MLP layers in various CTR prediction models. Extensive experiments on two academic datasets and one real industrial benchmark demonstrate that, with 3--5× model size reduction, our compressed models can achieve both faster inference and higher AUC than the uncompressed original models. Our code is at https://github.com/yuhao318/Atomic_Feature_Mimicking.

Graph convolutional network (GCN) has been successfully applied to capture global non-consecutive and long-distance semantic information for text classification. However, while GCN-based methods have shown promising results in offline evaluations, they commonly follow a seen-token-seen-document paradigm by constructing a fixed document-token graph and cannot make inferences on new documents. It is a challenge to deploy them in online systems to infer steaming text data. In this work, we present a continual GCN model (ContGCN) to generalize inferences from observed documents to unobserved documents. Concretely, we propose a new all-token-any-document paradigm to dynamically update the document-token graph in every batch during both the training and testing phases of an online system. Moreover, we design an occurrence memory module and a self-supervised contrastive learning objective to update ContGCN in a label-free manner. A 3-month A/B test on Huawei public opinion analysis system shows ContGCN achieves 8.86% performance gain compared with state-of-the-art methods. Offline experiments on five public datasets also show ContGCN can improve inference quality. The source code will be released at https://github.com/Jyonn/ContGCN.

Minyi Zhao, Lu Zhang, Yi Xu, Jiandong Ding, Jihong Guan, Shuigeng Zhou. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Weakly-supervised text classification has received much attention in recent years for it can alleviate the heavy burden of annotating massive data.Among them, keyword-driven methods are the mainstream where user-provided keywords are exploited to generate pseudolabels for unlabeled texts.However, existing methods treat keywords independently, thus ignore the correlation among them, which should be useful if properly exploited.In this paper, we propose a novel framework called ClassKG to explore keyword-keyword correlation on keyword graph by GNN.Our framework is an iterative process.In each iteration, we first construct a keyword graph, so the task of assigning pseudo labels is transformed to annotating keyword subgraphs.To improve the annotation quality, we introduce a self-supervised task to pretrain a subgraph annotator, and then finetune it.With the pseudo labels generated by the subgraph annotator, we then train a text classifier to classify the unlabeled texts.Finally, we re-extract keywords from the classified texts.Extensive experiments on both long-text and short-text datasets show that our method substantially outperforms the existing ones.

The scarcity of labeled data is a critical obstacle to deep learning. Semi-supervised learning (SSL) provides a promising way to leverage unlabeled data by pseudo labels. However, when the size of labeled data is very small (say a few labeled samples per class), SSL performs poorly and unstably, possibly due to the low quality of learned pseudo labels. In this paper, we propose a new SSL method called DP-SSL that adopts an innovative data programming (DP) scheme to generate probabilistic labels for unlabeled data. Different from existing DP methods that rely on human experts to provide initial labeling functions (LFs), we develop a multiple-choice learning~(MCL) based approach to automatically generate LFs from scratch in SSL style. With the noisy labels produced by the LFs, we design a label model to resolve the conflict and overlap among the noisy labels, and finally infer probabilistic labels for unlabeled samples. Extensive experiments on four standard SSL benchmarks show that DP-SSL can provide reliable labels for unlabeled data and achieve better classification performance on test sets than existing SSL methods, especially when only a small number of labeled samples are available. Concretely, for CIFAR-10 with only 40 labeled samples, DP-SSL achieves 93.82% annotation accuracy on unlabeled data and 93.46% classification accuracy on test data, which are higher than the SOTA results.

MiRNA are about 22nt long small noncoding RNAs that post transcriptionally regulate gene expression in animals, plants and protozoa. Confident identification of MiRNA-Target Interactions (MTI) is vital to understand their function. Currently, several integrated programs and databases are available for animal miRNAs, the mechanisms of which are significantly different from plant miRNAs. Here we present imiRTP, an integrated MTI prediction and analysis toolkit for Arabidopsis thaliana. It features two important functions: (i) combination of several effective plant miRNA target prediction methods provides a sufficiently large MTI candidate set, and (ii) different filters allow for an efficient selection of potential targets. The modularity of imiRTP enables the prediction of high quality targets on genome-wide scale.

BACKGROUND: MicroRNAs (miRNAs) are ~22 nt long integral elements responsible for post-transcriptional control of gene expressions. After the identification of thousands of miRNAs, the challenge is now to explore their specific biological functions. To this end, it will be greatly helpful to construct a reasonable organization of these miRNAs according to their homologous relationships. Given an established miRNA family system (e.g. the miRBase family organization), this paper addresses the problem of automatically and accurately classifying newly found miRNAs to their corresponding families by supervised learning techniques. Concretely, we propose an effective method, miRFam, which uses only primary information of pre-miRNAs or mature miRNAs and a multiclass SVM, to automatically classify miRNA genes. RESULTS: An existing miRNA family system prepared by miRBase was downloaded online. We first employed n-grams to extract features from known precursor sequences, and then trained a multiclass SVM classifier to classify new miRNAs (i.e. their families are unknown). Comparing with miRBase's sequence alignment and manual modification, our study shows that the application of machine learning techniques to miRNA family classification is a general and more effective approach. When the testing dataset contains more than 300 families (each of which holds no less than 5 members), the classification accuracy is around 98%. Even with the entire miRBase15 (1056 families and more than 650 of them hold less than 5 samples), the accuracy surprisingly reaches 90%. CONCLUSIONS: Based on experimental results, we argue that miRFam is suitable for application as an automated method of family classification, and it is an important supplementary tool to the existing alignment-based small non-coding RNA (sncRNA) classification methods, since it only requires primary sequence information. AVAILABILITY: The source code of miRFam, written in C++, is freely and publicly available at: http://admis.fudan.edu.cn/projects/miRFam.htm.

BACKGROUND: MicroRNAs (simply miRNAs) are derived from larger hairpin RNA precursors and play essential regular roles in both animals and plants. A number of computational methods for miRNA genes finding have been proposed in the past decade, yet the problem is far from being tackled, especially when considering the imbalance issue of known miRNAs and unidentified miRNAs, and the pre-miRNAs with multi-loops or higher minimum free energy (MFE). This paper presents a new computational approach, miRenSVM, for finding miRNA genes. Aiming at better prediction performance, an ensemble support vector machine (SVM) classifier is established to deal with the imbalance issue, and multi-loop features are included for identifying those pre-miRNAs with multi-loops. RESULTS: We collected a representative dataset, which contains 697 real miRNA precursors identified by experimental procedure and other computational methods, and 5428 pseudo ones from several datasets. Experiments showed that our miRenSVM achieved a 96.5% specificity and a 93.05% sensitivity on the dataset. Compared with the state-of-the-art approaches, miRenSVM obtained better prediction results. We also applied our method to predict 14 Homo sapiens pre-miRNAs and 13 Anopheles gambiae pre-miRNAs that first appeared in miRBase13.0, MiRenSVM got a 100% prediction rate. Furthermore, performance evaluation was conducted over 27 additional species in miRBase13.0, and 92.84% (4863/5238) animal pre-miRNAs were correctly identified by miRenSVM. CONCLUSION: MiRenSVM is an ensemble support vector machine (SVM) classification system for better detecting miRNA genes, especially those with multi-loop secondary structure.