Déjà Vu: The Importance of Time and Causality in Recommender Systems
by Justin Basilico (Netflix) and Yves Raimond (Netflix)
Time plays a key role in recommendation. Handling it properly is especially critical when using recommender systems in real-world applications, which may not be as clear when doing research with historical data. In this talk, we will discuss some of the important challenges of handling time in recommendation algorithms at Netflix. We will focus on challenges related to how our users, items, and systems all change over time. We will then discuss some strategies for tackling these challenges, which revolves around proper treatment of causality in our systems.
About the Speakers
Justin Basilico is a Research/Engineering Manager for Page Algorithms Engineering at Netflix. He leads an applied research team focused on developing the next generation of algorithms used to generate the Netflix homepage through machine learning, ranking, recommendation, and large-scale software engineering. He has also developed machine learning approaches that yielded significant improvements in the personalized ranking algorithms that drive the Netflix recommendation system. Prior to Netflix, he worked on machine learning in the Cognitive Systems group at Sandia National Laboratories. He is also the co-creator of the Cognitive Foundry, an open-source software library for building machine learning algorithms and applications.
Yves Raimond is a Research/Engineering Director at Netflix, where he leads the Search, Targeting & Recommendations Algorithm Engineering team: a mixed team of researchers and engineers building the next generation of Machine Learning algorithms used to drive the Netflix experience. Before that, he was a Lead Research Engineer in BBC R&D, working on information extraction from Multimedia content. He holds a PhD from Queen Mary, University of London.
Slides
https://www.slideshare.net/…/…-importance-of-time-and-causality-in-recommender-systems
Building Recommender Systems for Fashion
by Nick Landia (Dressipi)
Dressipi is a personalisation and style advice engine for women’s fashion. We work with some of the biggest retailers in the UK who have integrated our service into their site, and are currently expanding to the US and Australia. Since our launch in 2011 we have been helping millions of users find the clothes that they will love, buy and keep.
In this talk I will discuss the unique characteristics of the fashion domain and some of the most interesting challenges they pose for recommender systems. Fashion is inherently social and public: we dress not only for ourselves but also for the appropriateness of the environment we are in. Fashion recommendations must satisfy two sometimes competing objectives: identifying the user’s personal preference from their past behaviour and giving advice on what changes to their style would make them look better. Unlike other domains, recommendations should not be purely based on the user’s personal taste and past activity. They must also take public perception into account by being aware of fashion rules, outfit guidelines and current trends. To help overcome these challenges we have started gathering additional personal information about the users in questionnaires, if they wish to provide it. Examples of this include body shape, age, favourite colours, lifestyle etc. These additional data points allow for some exciting applications such as giving style advice and generating high quality recommendation reasons that are useful to the user.
About the Speaker
Nick Landia is the Chief Data Scientist at Dressipi where he has been working on recommendation algorithms for fashion for over three years. He is not only in charge of the recommendation and analytics teams, but also interacts with fashion domain experts on a regular basis to help guide the development of new approaches. The current areas of focus of his work are relating item features to user features, generating recommendation reasons and whole page optimisation. Previously he completed a PhD in the area of recommender systems at the University of Warwick, UK under the supervision of Dr. Sarabjot Singh Anand, and has co-founded two startups.
Slides
https://www.slideshare.net/NickLandia/building-recommender-systems-for-fashion
Boosting Recommender Systems with Deep Learning
by João Gomes (Farfetch)
Farfetch is a global fashion marketplace with a catalog that, at any time, has over 200 000 products spanning over 2000 brands from luxury boutiques all around the world. Finding the right product to the right customer is a challenge that, we, as Data Scientists working on the Recommendations team, are trying to solve using state-of-the art algorithms and disruptive technologies. Deep learning (DL) is an area of Machine Learning that has recently been brought to the spotlight for its breakthrough results across several domains. In this talk, we will provide an overview of some ongoing projects at Farfetch in which Deep Learning plays a major role. In particular, we will address how we use our extensive dataset of normalized product images together with state of the art convolutional neural networks, to extract features that can be used to provide visually relevant recommendations. We will also explain how, by leveraging data from thousands of hand-curated outfits by fashion experts within the company, we are able to model stylistic relationships and generalize a method for recommending complementary products using deep siamese neural networks.
These are merely a sample of problems we are tackling using DL at Farfetch. We believe that there are plenty of opportunities for application of these techniques to recommender systems and we look forward to discussing the potentials of this stream of research with the RecSys community.
About the Speaker
João Gomes is the Lead Data Scientist for the Recommendations team at Farfetch. Since joining
the company in 2015 he has helped design and develop new algorithms that power
personalized recommendations and discoverability across the multiple channels of the Farfetch
platform. His team’s mission is to leverage large amounts of data to know, understand and
ultimately inspire its customers. He holds a PhD in computational physics from the
University of Aveiro, Portugal.
Slides
https://recsys.acm.org/…/RecSys-17-Industry-Presentation-Farfetch.pdf
Personalization challenges in e-Learning
by Roberto Turrin (CloudAcademy)
In this talk we will discuss some of the common challenges in the e-learning industry that we are personally facing at CloudAcademy – an on-line continuous training platform for Cloud technologies – such as: heterogeneity of content to recommend, tracked user signals, and specific recommendation goals targeting learning objectives.
About the Speaker
Roberto is a well known researcher in the recommendation community, with a record of relevant scientific publications, some of them also presented at RecSys.
Roberto currently works as Head of Technology at CloudAcademy, an online learning platform focused on cloud technologies, where he leads the research team.
Slides
https://www.slideshare.net/robertoturrin/personalization-challenges-in-elearning
Personalized Job Recommendation System at LinkedIn: Practical Challenges and Lessons Learned
by Benjamin Le (LinkedIn)
In this talk we overview the overall system design and architecture of LinkedIn Job Recommendations, the challenges encountered in practice, and the lessons learned from the production deployment of these systems at LinkedIn. Specifically, we will highlight how candidate selection of documents, personalization of member’s recommendations, and optimizion of the job marketplace are formulated and addressed at LinkedIn. By presenting our experiences of applying techniques at the intersection of recommender systems, information retrieval, machine learning, and statistical modeling in a large-scale industrial setting and highlighting the open problems, we hope to stimulate further research and collaborations with the RecSys community.
About the Speaker
Benjamin Le is a Senior Software Engineer at LinkedIn. His contributions to jobs relevance include candidate selection in job search through multi-pass ranking and deploying generalized linear mixed models for reducing poor job recommendation impressions. Currently, he is focusing on the deployment of deep and wide models for job recommendations. He is a recent graduate of the University of California, Berkeley, where he received both his Master of Engineering and Bachelor of Science Degrees.
Slides
https://www.slideshare.net/…/personalized-job-recommendation-system-at-linkedin…
