Recommender systems should adapt to user interests as the latter evolve. A prevalent cause for the evolution of user interests is the influence of their social circle. In general, when the interests are not known, online algorithms that explore the recommendation space while also exploiting observed preferences are preferable. We present online recommendation algorithms rooted in the linear multi-armed bandit literature. Our bandit algorithms are tailored precisely to recommendation scenarios where user interests evolve under social influence. In particular, we show that our adaptations of the classic LinREL and Thompson Sampling algorithms maintain the same asymptotic regret bounds as in the non-social case. We validate our approach experimentally using both synthetic and real datasets.

There is no doubt that on the whole, the economic impacts from the lockdown and pandemic will be devastating. But while most leisure activities were throttled by the lockdown, others thrived -- just ask any of your friends that did Yoga With Adrienne (probably the same mates that brew their own kombucha). Tinder and Bumble usage alone spiked by over 20%, with Tinder registering 3 billion swipes on 28 March alone. However, the pandemic only accelerated a trend that was already in full force: finding love via apps. "Met online" is now the most common way that people report finding their significant other, streets ahead of boring old classics like "met in church" or "met in the neighbourhood". While there are a range of massively popular dating apps, including Bumble and Grindr, Tinder continues to be the most popular platform by a significant margin.

President Donald Trump said Saturday he has approved a deal in principle in which Oracle and Walmart will partner with the viral video-sharing app TikTok in the U.S., allowing the popular app to avoid a shutdown. "I have given the deal my blessing -- if they get it done that's great, if they don't that's okay too," Trump told reporters on the White House South Lawn before departing for North Carolina. "I approved the deal in concept." The U.S. Department of Commerce announced it would delay the prohibition of U.S. transactions with TikTok until next Sunday. Shortly after Trump's comments, Oracle announced it was chosen as TikTok's secure cloud provider and will become a minority investor with a 12.5% stake.

I have been on a 30-day challenge to improve my knowledge of Artificial Intelligence (AI), to understand how it works and how it impacts our lives, and this section talks about how not only have we already integrated it in our everyday lives, but in some cases already love it and depend on it. In this fifth section, we tackle "AI in Application." Exploring where AI is prevalent and the data that is being collected already is not surprising but it is humbling how much it has already penetrated our lives and how much we depend on it. Recently, a friend of mine named her baby Sirius. For those that love the Harry Potter books, the immediate connection is to Sirius Black, so of course being a Harry Potter fan I instantly loved it.

In this episode of the Data Exchange I speak with Denise Gosnell, Chief Data Officer at DataStax1. Denise is also the co-author of the new book, The Practitioner's Guide to Graph Data, which covers foundational tools and techniques needed to utilize graph technologies in production applications. This conversation is a great introduction to what has become an important class of technologies and tools. Graph technologies are used to power a wide array of applications, including recommendation engines, fraud detection systems, identity and access management, search, and many other use cases. Denise provides a set of practical recommendations and advice for developers who are interested in unlocking the power of large graphs.

Smart watches and fitness trackers are only the first signs of a world that will enfold us in a subtle but ubiquitous web of connectivity. We have digital assistants that can make a restaurant reservation, lock the garage door, or check whether our laundry is dry when asked. We have tablets with cameras that automatically track us as we move around to make our video calls feel more like being there. We even have home automation hubs that use subsonic sound waves to locate us in a room and make their displays more readable from a distance. The next logical step will be not needing to interact deliberately with our devices at all.

In the early 2000s, Songza implemented a manual music recommendation system for its listeners, where a team of music experts and curators would create playlists. But these recommendations were not objective, as they were dependent on the personal taste of the curators. It was an average experience for listeners, with a fair share of hits and misses, because it was impossible to make a playlist which catered to the varied tastes of a diverse set of people. The technology and the data did not exist back then to build a playlist that would be personalised to the taste of each individual listener. Along came Spotify a few years later, offering a highly personalised weekly playlist called Discover Weekly that quickly became one of their flagship offerings.

Simulating online recommender system performance is notoriously difficult and the discrepancy between the online and offline behaviors is typically not accounted for in offline evaluations. Recommender systems research tends to evaluate model performance on randomly sampled targets, yet the same systems are later used to predict user behavior sequentially from a fixed point in time. This disparity permits weaknesses to go unnoticed until the model is deployed in a production setting. We first demonstrate how omitting temporal context when evaluating recommender system performance leads to false confidence. To overcome this, we propose an offline evaluation protocol modeling the real-life use-case that simultaneously accounts for temporal context. Next, we propose a training procedure to further embed the temporal context in existing models: we introduce it in a multi-objective approach to traditionally time-unaware recommender systems. We confirm the advantage of adding a temporal objective via the proposed evaluation protocol. Finally, we validate that the Pareto Fronts obtained with the added objective dominate those produced by state-of-the-art models that are only optimized for accuracy on three real-world publicly available datasets. The results show that including our temporal objective can improve recall@20 by up to 20%.

This week was jam-packed with new product announcements but we also got a couple of good deals, too. Apple's AirPods Pro remain at their lowest price ever -- only $199 -- and the latest MacBook Air is $100 off. Those in need of a new smartphone can grab the latest from Samsung for $200 less and the solid Beats Solo Pro headphones are down to $199. Here are the best deals from this week that you can still buy today. You can get AirPods Pro for their lowest price ever, only $199, on Amazon, Walmart or Staples.

Tinder is a mobile dating app that can help you find singles in the local area. "Swipe right if you like her, Swipe left if you don't" is a linchpin to the company's success, and the format has been duplicated by numerous contemporaries. Tinder was first launched as a location-based dating app in 2012 within incubator Hatch Labs and join a venture between IAC and Xtreme Labs and now it's one of the most popular dating apps in the US with about 1.7 Billion swipes per day. Tinder has employed the freemium business model to earn revenue. It went from a "location-based" dating app to a global dating app that is present in 190 countries in less than 8 years.