Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering
By Ruining He, Julian McAuley. The Web Conference, 2016
2,362 citations recorded by the source corpus.
Building a successful recommender system depends on understanding both the dimensions of people's preferences as well as their dynamics. In certain domains, such as fashion, modeling such preferences can be incredibly difficult, due to the need to simultaneously model the visual appearance of products as well as their evolution over time. The subtle semantics and non-linear dynamics of fashion evolution raise unique challenges especially considering the sparsity and large scale of the underlying datasets. In this paper we build novel models for the One-Class Collaborative Filtering setting, where our goal is to estimate users' fashion-aware personalized ranking functions based on their past feedback. To uncover the complex and evolving visual factors that people consider when evaluating products, our method combines high-level visual features extracted from a deep convolutional neural network, users' past feedback, as well as evolving trends within the community. Experimentally we evaluate our method on two large real-world datasets from Amazon.com, where we show it to outperform state-of-the-art personalized ranking measures, and also use it to visualize the high-level fashion trends across the 11-year span of our dataset.
Subfields: Computer Vision and Pattern Recognition, Information Systems, Marketing, Transportation
View Paper: https://arxiv.org/abs/1602.01585
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Xiaogang Wang
- Ming-Hsuan Yang
- Bernt Schiele
- Jitendra Malik
- Yi Yang
In Information Systems
In Marketing
In Transportation
Citations in the corpus
- Pretrained Transformers Improve Out-of-Distribution Robustness (2020, background): “Figure 6 shows the Amazon results for BERT Base, Table 5 shows the MNLI results, and”
- Refining Language Models with Compositional Explanations (2021, methodology): “For sentiment analysis, we first train on AmazonMusic [16], and apply the model to the Stanford Sentiment Treebank-2 (SST-2) dataset [40].”
- WT5?! Training Text-to-Text Models to Explain their Predictions (2020, methodology): “To test this transfer, we used the Amazon Reviews dataset (He & McAuley, 2016; McAuley et al., 2015) which contains reviews for a wide variety of products across diverse categories.”
- Don’t Stop Pretraining: Adapt Language Models to Domains and Tasks (2020, neutral): “75M AMAZON reviews (He and McAuley, 2016) 2.”
- Deep Interest Network for Click-Through Rate Prediction (2017, methodology): “is a widely used metric[8]. It measures the goodness of order by ranking all the ads with predicted CTR, including intra-user and inter-user orders. An variation of user weighted AUC is introduced in [7, 13] which measures the goodness of intra-user order by averaging AUC over users and is shown to ”
- ELLE: Efficient Lifelong Pre-training for Emerging Data (2022, background): “, 2019), AMAZON REVIEWS (REV) (He and McAuley, 2016), BIOMEDICAL PAPERS (BIO) (Lo et al.”
- Well-Read Students Learn Better: The Impact of Student Initialization on Knowledge Distillation (2019, methodology): “We reaffirm this statement through experiments on Amazon Book Reviews in Figure 5, given that Amazon Book Reviews have the biggest transfer set.”
- FLEX: Unifying Evaluation for Few-Shot NLP (2021, background): “• Class Transfer: FewRel [28], HuffPost [46], Amazon [30], 20News [38], and Reuters [41] take part in meta-training and meta-testing but with different classes.”
- BERT Post-Training for Review Reading Comprehension and Aspect-based Sentiment Analysis (2019, methodology): “For domain knowledge post-training, we use Amazon laptop reviews (He and McAuley, 2016) and Yelp Dataset Challenge reviews8.”
- Open-world Learning and Application to Product Classification (2018, methodology): “We leverage the huge amount of product descriptions from the Amazon Datasets [14] and form the OWL task as the following.”
- Politeness Transfer: A Tag and Generate Approach (2020, methodology): “We also use the Amazon dataset of product reviews (He and McAuley, 2016).”
- Style Transfer in Text: Exploration and Evaluation (2017, methodology): “Positive-Negative Review Dataset This dataset contains Amazon product reviews published by He and McAuley (2016).”
- UltraGCN: Ultra Simplification of Graph Convolutional Networks for Recommendation (2021, methodology): “For example, in our experiments, three-layer LightGCN takes more than 700 epochs to converge to its best result on the AmazonBooks dataset [9], which would be unacceptable in an industrial setting.”
- Neural Graph Collaborative Filtering (2019, methodology): “sses like restaurants and bars are viewed as the items. We use the same 10-core setting in order to ensure data quality. Amazon-book: Amazon-review is a widely used dataset for product recommendation [9]. We select Amazon-book from the collection. Similarly, we use the 10-core setting to ensure that”
- Data Selection for Language Models via Importance Resampling (2023, methodology): “For manual curation (DAPT), we attempted to download the datasets used in the paper (RealNews [91], S2ORC [49], and Amazon reviews [25]).”
- Multi-Interest Network with Dynamic Routing for Recommendation at Tmall (2019, background): “One is Amazon Books1 provided by [10, 20], representing one of the most widely-used public dataset for e-commerce recommendations.”