A5100699702- Recommender Systems and Techniques
- Advanced Bandit Algorithms Research
- Topic Modeling
- Natural Language Processing Techniques
- Neural Networks and Applications
- Data Stream Mining Techniques
- Machine Learning and Algorithms
- Reinforcement Learning in Robotics
- Complex Network Analysis Techniques
- Smart Grid Energy Management
- Domain Adaptation and Few-Shot Learning
- Machine Learning and Data Classification
- Machine Learning in Healthcare
- Model Reduction and Neural Networks
- Expert finding and Q&A systems
- Image Retrieval and Classification Techniques
- Text and Document Classification Technologies
- Advanced Graph Neural Networks
- Speech Recognition and Synthesis
- Generative Adversarial Networks and Image Synthesis
- Data Management and Algorithms
- Anomaly Detection Techniques and Applications
- Opinion Dynamics and Social Influence
- Algorithms and Data Compression
- Sentiment Analysis and Opinion Mining
Google (United States)
2018-2024
Huawei Technologies (China)
2024
Anhui University
2023
Anqing Normal University
2023
Amazon (United States)
2013-2021
Criteo (France)
2014-2017
Washington University in St. Louis
2008-2013
University of Washington
2012
East China University of Science and Technology
2010
Stacked denoising autoencoders (SDAs) have been successfully used to learn new representations for domain adaptation. Recently, they attained record accuracy on standard benchmark tasks of sentiment analysis across different text domains. SDAs robust data by reconstruction, recovering original features from that are artificially corrupted with noise. In this paper, we propose marginalized SDA (mSDA) addresses two crucial limitations SDAs: high computational cost and lack scalability...
Industrial recommender systems deal with extremely large action spaces -- many millions of items to recommend. Moreover, they need serve billions users, who are unique at any point in time, making a complex user state space. Luckily, huge quantities logged implicit feedback (e.g., clicks, dwell time) available for learning. Learning from the is however subject biases caused by only observing on recommendations selected previous versions recommender. In this work, we present general recipe...
Recurrent neural networks have gained widespread use in modeling sequential data. Learning long-term dependencies using these models remains difficult though, due to exploding or vanishing gradients. In this paper, we draw connections between recurrent and ordinary differential equations. A special form of called the AntisymmetricRNN is proposed under theoretical framework, which able capture thanks stability property its underlying equation. Existing approaches improving RNN trainability...
The large-scale production of edible mushrooms typically requires the use greenhouses, as greenhouse environment significantly affects growth mushrooms. It is crucial to effectively predict temperature, humidity, and carbon dioxide fluctuations within mushroom for determining environmental stress pre-regulation To address nonlinearity, temporal dynamics, strong coupling environment, a prediction model based on combination attention mechanism, convolutional neural network, long short-term...
Machine learning algorithms have successfully entered industry through many real-world applications (e.g., search engines and product recommendations). In these applications, the test-time CPU cost must be budgeted accounted for. this paper, we examine two main components of cost, classifier evaluation feature extraction show how to balance costs with accuracy. Since computation required for dominates a in settings, develop efficiently performance cost. Our first contribution describes...
We present an efficient document representation learning framework, Document Vector through Corruption (Doc2VecC). Doc2VecC represents each as a simple average of word embeddings. It ensures generated such captures the semantic meanings during learning. A corruption model is included, which introduces data-dependent regularization that favors informative or rare words while forcing embeddings common and non-discriminative ones to be close zero. produces significantly better than Word2Vec....
Many real-world recommender systems need to be highly scalable: matching millions of items with billions users, milliseconds latency. The scalability requirement has led widely used two-stage systems, consisting efficient candidate generation model(s) in the first stage and a more powerful ranking model second stage.
Over the years we have seen recommender systems shifting focus from optimizing short-term engagement toward improving long-term user experience on platforms. While defining good is still an active research area, one specific aspect of improved here, which revisiting platform. These long term outcomes however are much harder to optimize due sparsity in observing these events and low signal-to-noise ratio (weak connection) between a single recommendation. To address challenges, propose...
The reasoning and generalization capabilities of LLMs can help us better understand user preferences item characteristics, offering exciting prospects to enhance recommendation systems. Though effective while user-item interactions are abundant, conventional systems struggle recommend cold-start items without historical interactions. To address this, we propose utilizing as data augmenters bridge the knowledge gap on during training. We employ infer for based textual description behaviors...
Understanding temporal dynamics has proved to be highly valuable for accurate recommendation. Sequential recommenders have been successful in modeling the of users and items over time. However, while different model architectures excel at capturing various ranges or dynamics, distinct application contexts require adapting diverse behaviors.
Reinforcement Learning (RL) techniques have been sought after as the next-generation tools to further advance field of recommendation research. Different from classic applications RL, recommender agents, especially those deployed on commercial platforms, operate in extremely large state and action spaces, serving a dynamic user base order billions, long-tail item corpus millions or billions. The (positive) feedback available train such agents is scarce retrospect. Improving sample efficiency...
Industrial recommendation platforms are increasingly concerned with how to make recommendations that cause users enjoy their long term experience on the platform. Reinforcement learning emerged naturally as an appealing approach for its promise in 1) combating feedback loop effect resulted from myopic system behaviors; and 2) sequential planning optimize outcome. Scaling RL algorithms production recommender systems serving billions of contents, however remain challenging. Sample inefficiency...
Effective exploration is believed to positively influence the long-term user experience on recommendation platforms. Determining its exact benefits, however, has been challenging. Regular A/B tests often measure neutral or even negative engagement metrics while failing capture benefits. We here introduce new experiment designs formally quantify value of by examining effects content corpus, and connecting corpus growth from real-world experiments. Once established values exploration, we...
Stacked denoising autoencoders (SDAs) have been successfully used to learn new representations for domain adaptation. They attained record accuracy on standard benchmark tasks of sentiment analysis across different text domains. SDAs robust data by reconstruction, recovering original features from that are artificially corrupted with noise. In this paper, we propose marginalized Linear Denoising Autoencoder (mSLDA) addresses two crucial limitations SDAs: high computational cost and lack...