Title: Cross Modal Cooking Recipe Retrieval
Speaker: Chen Jingjing （陈静静博士）
Place: N3 332（N3楼332会议室）
Time: 24th August 2018, 10:00-10:30 （2018年8月24日， 周五）
Host: Nie Liqiang （聂礼强）
Abstract:
In social media users like to share food pictures. One intelligent feature, potentially attractive to amateur chefs, is the recommendation of recipe along with food. Having this feature, unfortunately, is still technically challenging. First, the current technology in food recognition can only scale up to a few hundreds of categories, which are yet to be practical for recognizing tens of thousands of food categories. Second, even one food category can have variants of recipes that differ in ingredient composition. Finding the best-match recipe requires knowledge of ingredients, which is a fine-grained recognition problem. This talk will share you my recent work on cross-modal recipe retrieval, including a multi-task deep CNN architecture that explores the mutual but fuzzy relationship between food and ingredients for simultaneous learning/recognition; and a stacked attention model that learns joint-embedding feature for image-recipe similar learning.
Biography:
Chen Jingjing received her Ph.D. degree from City University of Hong Kong, supervised by Prof. Chong-Wah Ngo. She received her master degree from Tianjin University in 2014 and got her Bachelor degree from Wuhan University of Technology, in 2011. Her research interest lies in the areas of diet tracking and nutrition estimation based on multi-modal processing of food images. Currently, she is focusing on cross-modal recipe retrieval and food image recognition. Various parts of her work have been published in ACM Multimedia, Multimedia Modeling. She won the best student paper award in ACM Multimedia 2016 and Multimedia Modeling 2017, respectively. Currently, she is a research fellow in National University of Singapore, supervised by Professor Chua.

Title: Temporal Relational Model for Asset Price Estimation
Speaker: Feng Fuli （冯福利）
Place: N3 332 （N3楼332会议室）
Time: 24th August 2018, 10:30-11:00 （2018年8月24日，周五）
Host: Nie Liqiang （聂礼强）
Asset price estimation is widely used in financial investment to forecast financial markets and facilitate investment decisions. Traditional solutions for asset price estimation are based on financial models fed with core market data. With the recent success of big data and machine learning, machine learning models, especially deep neural networks, taking heterogeneous data including both traditional market data and large-scale alternative data as input, have become promising choices for asset price estimation. However, the existing solutions tend to achieve suboptimal prediction performance since ignoring the rich explicit in financial domain. As such, we aim to develop more promising neural network-based solutions for asset price estimation via incorporating relational financial data. Taking stock price estimation as an example, we contribute a new deep learning solution, named Relational Stock Ranking (RSR). Our RSR method advances over existing solutions captures the stock relations in a time-sensitive manner via a new component in neural network modeling, named Temporal Graph Convolution, which jointly models the temporal evolution and relation network of stocks. Extensive experiments demonstrate the superiority of our RSR method, which outperforms state-of-the-art stock price estimation solutions, and achieves an average return ratio of 98% and 71% on NYSE and NASDAQ, respectively.
Biography:
Feng Fuli is a Ph.D. student in the School of Computing, National University of Singapore. He received the B.E. degree in School of Computer Science and Engineering from Baihang University, Beijing, in 2015. His research interests include information retrieval, data mining, and multi-media processing. He has over 10 publications appeared in several top conferences such as SIGIR, WWW, and MM. His work on Bayesian Personalized Ranking has received the Best Poster Award of WWW 2018. Moreover, he has been served as the PC member and external reviewer for several top conferences including SIGIR, ACL, KDD, IJCAI, AAAI, WSDM etc.

Title: TEM: Tree-enhanced Embedding Model for Explainable Recommendation
Speaker: Wang Xiang （王翔）
Place: N3 332 （N3楼332会议室）
Time: 24th August 2018, 11:00 （2018年8月24日，周五）
Host: Nie Liqiang （聂礼强）
Abstract:
While collaborative filtering is the dominant technique in personalized recommendation, it models user-item interactions only and cannot provide concrete reasons for a recommendation. Meanwhile, the rich side information affiliated with user-item interactions, providing valuable evidence, has not been fully explored in providing explanations. On the technical side, embedding-based methods, such as Wide & Deep and neural factorization machines, provide state-of-the-art recommendation performance. However, they work like a black-box, for which the reasons underlying a prediction cannot be explicitly presented. In this work, we propose a novel solution named Tree-enhanced Embedding Method that combines the strengths of embedding-based and tree-based models. We first employ a tree-based model to learn explicit decision rules (aka. cross features) from the rich side information. We next design an embedding model that can incorporate explicit cross features and generalize to unseen cross features on user ID and item ID. The core of our embedding method is an easy-to-interpret attention network, making the recommendation process fully transparent and explainable. We conduct experiments on two datasets of tourist attraction and restaurant recommendation, demonstrating the superior performance and explainability of our solution.
Xiang Wang is currently a Ph.D. student in the School of Computing, National University of Singapore. He received the B.E. degree in computer science and engineering from Baihang University, Beijing, in 2014. His research interests include information retrieval, social media analysis, and grouping discovery and profiling from social media. Various parts of his work have been published in top forums, such as ACM SIGIR, WWW, TOIS, and MM. He has been served as reviewers for various conferences, such as MMM and WebScience.