Learning and Transferring IDs Representation in E-commerce

Zhao, Kui; Li, Yuechuan; Shuai, Zhaoqian; Yang, Cheng

doi:10.1145/3219819.3219855

Cited by 21 publications

(16 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the application of deep learning in RS, some internet companies adopted neural networks to address the cold-start problem [38]- [40]. Alibaba uses the embedding of side information in new items to replace the embedding of the items to calculate the similarity between items [38], [39]. For the cold-start problem of housing options, Airbnb selects the three closest houses of the same type (within a radius of 10 miles) and uses the average value of the embedding of the three houses as the embedding of the new house [40].…”

Section: Recent Deep Learning Research On the Cold-start Problemmentioning

confidence: 99%

Facing Cold-Start: A Live TV Recommender System Based on Neural Networks

Zhu

Guo

et al. 2020

IEEE Access

View full text Add to dashboard Cite

With the increase in the number of live TV channels, audiences must spend increasing amounts of time and energy deciding which shows to watch; this problem is called information overload, and recommender systems (RSs) are effective methods for addressing such problems. Due to the high update rates and low replay rates of TV programs, the item cold-start problem is prominent, and this problem seriously affects the effectiveness of the recommender and limits the application of recommendation algorithms for live TV. To solve this problem better, RSs must consider information in addition to the time slot strategy, which relies on experience. At present, no methods make good use of viewing behavior records. Therefore, in this paper, we proposed a viewing environment model called DeepTV that considers viewing behavior records and electronic program guides and includes a feature generation process and a model construction process. In the feature generation process, we defined seven key features by clustering viewing time, distinguishing positive and negative feedback, capturing continuous viewing preference and introducing the remaining time proportion of candidate programs. We normalize the continuous features and add powers of them. In the model construction process, we regard the live TV recommendation task as a classification problem and fuse the above features by using a neural network. Finally, experiments on industrial datasets show that the proposed model significantly outperforms baseline algorithms.

show abstract

Section: Recent Deep Learning Research On the Cold-start Problemmentioning

confidence: 99%

Facing Cold-Start: A Live TV Recommender System Based on Neural Networks

Zhu

Guo

et al. 2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Word2vec is a popular model for learning word representations that has since found a wide range of additional uses [2,21,8,22,5,20,9,6,17]. Owing to its robustness, simplicity, and efficiency, Word2vec is a valuable component of many recommender systems where it is used for benchmarking, candidate generation and transfer learning [2,21,8,22,5,20,9,16]. It is common for the default parameters given in [12,18] to be taken (e.g.…”

Section: Introductionmentioning

confidence: 99%

Tuning Word2vec for Large Scale Recommendation Systems

Chamberlain¹,

Rossi²,

Shiebler

et al. 2020

Fourteenth ACM Conference on Recommender Systems

View full text Add to dashboard Cite

Word2vec is a powerful machine learning tool that emerged from Natural Language Processing (NLP) and is now applied in multiple domains, including recommender systems, forecasting, and network analysis. As Word2vec is often used off the shelf, we address the question of whether the default hyperparameters are suitable for recommender systems. The answer is emphatically no. In this paper, we first elucidate the importance of hyperparameter optimization and show that unconstrained optimization yields an average 221% improvement in hit rate over the default parameters. However, unconstrained optimization leads to hyperparameter settings that are very expensive and not feasible for large scale recommendation tasks. To this end, we demonstrate 138% average improvement in hit rate with a runtime budget-constrained hyperparameter optimization. Furthermore, to make hyperparameter optimization applicable for large scale recommendation problems where the target dataset is too large to search over, we investigate generalizing hyperparameters settings from samples. We show that applying constrained hyperparameter optimization using only a 10% sample of the data still yields a 91% average improvement in hit rate over the default parameters when applied to the full datasets. Finally, we apply hyperparameters learned using our method of constrained optimization on a sample to the Who To Follow recommendation service at Twitter and are able to increase follow rates by 15%.

show abstract

“…Such a framework will reduce the cost of re-building and recalibrating the learning models due to changes of distribution and input space features, which is known as 'transfer learning'. Transfer learning is useful in many real world applications, such as natural language processing(NLP) (Han & Eisenstein 2019, Kim, Gao & Ney 2019, medical and clinical analysis (Christodoulidis, et al 2016, Uran, et al 2019), E-commerce (Zhao, Li, Shuai, & Yang, 2018) and acoustic recognition (Gharib, Drossos, Çakir, Serdyuk, & Virtanen, 2018). In our study, we aim to address the distribution divergence in the a large dataset consisting of several subsets by transfer learning.…”

Section: Introductionmentioning

confidence: 99%