Farewell 2017, a year in which marketing technology really took off and where both consumers and marketers alike started to experience the potential of artificial intelligence in our daily personal and professional lives. Thanks to our new voice-operated companions Alexa and Siri, and advanced analytics tools based on machine learning becoming increasingly accessible, we all caught a glimpse of the exciting future driven by AI. Certainly, interest in AI among marketers this past year was the highest it has ever been. It was difficult not to stumble upon some discussion about AI and marketing in every conference, blog, trade press article or pitch. That's a very healthy interest, and we should encourage it into the new year.

Both of the major smart home platforms have a long-running problem with "discoverability": it's very hard to let users know what their devices can do, particularly if they're always improving thanks to rapid software updates. Amazon and Google are constantly experimenting with ways to get around this, but so far they have been timid. Amazon sends a weekly email, while Google includes some tips in its app. Expect to see them be bolder, particularly as powerful rivals such as Apple appear on the scene with worse AI but better sound. So don't be surprised if your Google Home or Amazon Echo begin to talk back, rather than simply following commands.

OkCupid is facing a lot of backlash for ditching usernames in favor of real names. As TechCrunch noted, its app's Google Play and App Store pages are flooded with one-star reviews posted over the past month, complaining about the features the service has recently changed or killed. The reviews talked about how the forced transparency of the new name requirement could compromise people's privacy and security, since the service now also matches users with others nearby based on their shared interests. Those with stalkery tendencies might take advantage of these new offerings. A lot of them also discussed concerns that Match.com is simply turning OkCupid (which it owns) into another Tinder (which it also owns).

LG is releasing a new line of premium audio devices in 2018. Ahead of the products' release, the South Korea giant has decided to introduce them via its online newsroom. Among the home speakers that LG is launching are the ThinQ Speaker and the SK10Y soundbar. LG's first premium audio device for 2018 is the ThinQ Speaker which features Google Assistant. Aside from giving users access to a digital assistant, the audio product is also built to offer high-quality sound experience.

At a high level, Artificial Intelligence (AI) is a branch of computer science that makes machines imitate intelligent human behavior, simulating (and often exceeding) human performance. AI has finally emerged as the future, after unfulfilled hype that goes back to the 1950s, due to developments such as the availability of an immense amount of data, the open-sourcing of ML algorithm development, and advances in high-density parallel processing infrastructure. In fact, IBM now believes the technology solutions market for AI amounts to a staggering $2 trillion over the next decade. Data is the new oil, and 90% of data in the world right now has been created in the last 2 years alone. The power of data has actually lagged the technical capability to monetize it efficiently and effectively, in a world where the use of data is moving from a competitive advantage to a requirement to compete.

Low-rank matrix approximation (LRMA) methods have achieved excellent accuracy among today's collaborative filtering (CF) methods. In existing LRMA methods, the rank of user/item feature matrices is typically fixed, i.e., the same rank is adopted to describe all users/items. However, our studies show that submatrices with different ranks could coexist in the same user-item rating matrix, so that approximations with fixed ranks cannot perfectly describe the internal structures of the rating matrix, therefore leading to inferior recommendation accuracy. In this paper, a mixture-rank matrix approximation (MRMA) method is proposed, in which user-item ratings can be characterized by a mixture of LRMA models with different ranks. Meanwhile, a learning algorithm capitalizing on iterated condition modes is proposed to tackle the non-convex optimization problem pertaining to MRMA. Experimental studies on MovieLens and Netflix datasets demonstrate that MRMA can outperform six state-of-the-art LRMA-based CF methods in terms of recommendation accuracy.

Singular values of a data in a matrix form provide insights on the structure of the data, the effective dimensionality, and the choice of hyper-parameters on higher-level data analysis tools. However, in many practical applications such as collaborative filtering and network analysis, we only get a partial observation. Under such scenarios, we consider the fundamental problem of recovering various spectral properties of the underlying matrix from a sampling of its entries. We propose a framework of first estimating the Schatten $k$-norms of a matrix for several values of $k$, and using these as surrogates for estimating spectral properties of interest, such as the spectrum itself or the rank. This paper focuses on the technical challenges in accurately estimating the Schatten norms from a sampling of a matrix. We introduce a novel unbiased estimator based on counting small structures in a graph and provide guarantees that match its empirical performances. Our theoretical analysis shows that Schatten norms can be recovered accurately from strictly smaller number of samples compared to what is needed to recover the underlying low-rank matrix. Numerical experiments suggest that we significantly improve upon a competing approach of using matrix completion methods.

Latent models have become the default choice for recommender systems due to their performance and scalability. However, research in this area has primarily focused on modeling user-item interactions, and few latent models have been developed for cold start. Deep learning has recently achieved remarkable success showing excellent results for diverse input types. Inspired by these results we propose a neural network based latent model called DropoutNet to address the cold start problem in recommender systems. Unlike existing approaches that incorporate additional content-based objective terms, we instead focus on the optimization and show that neural network models can be explicitly trained for cold start through dropout. Our model can be applied on top of any existing latent model effectively providing cold start capabilities, and full power of deep architectures. Empirically we demonstrate state-of-the-art accuracy on publicly available benchmarks. Code is available at https://github.com/layer6ai-labs/DropoutNet.

Word embeddings are an effective tool to analyze language. They have been recently extended to model other types of data beyond text, such as items in recommendation systems. Embedding models consider the probability of a target observation (a word or an item) conditioned on the elements in the context (other words or items). In this paper, we show that conditioning on all the elements in the context is not optimal. Instead, we model the probability of the target conditioned on a learned subset of the elements in the context. We use amortized variational inference to automatically choose this subset. Compared to standard embedding models, this method improves predictions and the quality of the embeddings.

The sparse matrix estimation problem consists of estimating the distribution of an $n\times n$ matrix $Y$, from a sparsely observed single instance of this matrix where the entries of $Y$ are independent random variables. This captures a wide array of problems; special instances include matrix completion in the context of recommendation systems, graphon estimation, and community detection in (mixed membership) stochastic block models. Inspired by classical collaborative filtering for recommendation systems, we propose a novel iterative, collaborative filtering-style algorithm for matrix estimation in this generic setting. We show that the mean squared error (MSE) of our estimator converges to $0$ at the rate of $O(d^2 (pn)^{-2/5})$ as long as $\omega(d^5 n)$ random entries from a total of $n^2$ entries of $Y$ are observed (uniformly sampled), $\E[Y]$ has rank $d$, and the entries of $Y$ have bounded support. The maximum squared error across all entries converges to $0$ with high probability as long as we observe a little more, $\Omega(d^5 n \ln^5(n))$ entries. Our results are the best known sample complexity results in this generality.