In this paper we propose a model that combines the strengths of RNNs and SGVB: the Variational Recurrent Auto-Encoder (VRAE). Such a model can be used for efficient, large scale unsupervised learning on time series data, mapping the time series data to a latent vector representation. The model is generative, such that data can be generated from samples of the latent space. An important contribution of this work is that the model can make use of unlabeled data in order to facilitate supervised training of RNNs by initialising the weights and network state.

Apple wants to control every aspect of your life -- as long as you choose to let it. The company on Monday showed off the newest updates to the software running on its iPhones and iPads, Macs and Apple Watches. One of the key characteristics of Apple's operating system releases over the past couple years -- including iOS 9 and Mac OS X 10.11 El Capitan revealed Monday at its developer conference -- is how well the software makes its devices work together. Last year's debut of Continuity and Handoff tied the iPhone and Mac computer more closely together, letting people start an email on their smartphone and finish it on their computer or even answer phone calls on their Macs. This year, Apple showed more ways it will extend its reach into our homes and cars, as well as how we listen to music and how we pay for goods.

Apple on Monday served up a veritable smorgasbord of digital delights for its fans, unveiling at its annual developers conference upgrades to its mobile and desktop software, showing off a gussied-up Siri with a new bag of tricks, and firing a shot over Spotify's bow with its new streaming Apple Music subscription service. "This is a truly revolutionary music service," Eddy Cue, Apple's senior vice president of Internet software and services, told the crowd of several thousand developers, designers and product managers at the 26th Worldwide Developers Conference, the annual Apple love fest at Moscone Center in San Francisco. "Apple Music will bring you all of your music all in one place." Revealed toward the end of a nearly three-hour extravaganza, the music feature was clearly Apple's rabbit out of a hat. It had been widely expected for months, ever since May last year when Apple bought subscription streaming music service Beats Music, and Beats Electronics, which makes the popular Beats headphones, speakers and audio software.

Apple on Monday served up a veritable smorgasbord of digital delights for its fans, unveiling at its annual developers conference upgrades to its mobile and desktop software, showing off a gussied-up Siri with a new bag of tricks, and firing a shot over Spotify's bow with its new streaming Apple Music subscription service. "This is a truly revolutionary music service," Eddy Cue, Apple's senior vice president of Internet software and services, told the crowd of several thousand developers, designers and product managers at the 26th Worldwide Developers Conference, the annual Apple love fest at Moscone Center in San Francisco. "Apple Music will bring you all of your music all in one place." Revealed toward the end of a nearly three-hour extravaganza, the music feature was clearly Apple's rabbit out of a hat. It had been widely expected for months, ever since May last year when Apple bought subscription streaming music service Beats Music, and Beats Electronics, which makes the popular Beats headphones, speakers and audio software.

During Apple's annual developers keynote at WWDC, Senior Vice President of Software Engineering Craig Federighi announced the company's latest mobile operating system, iOS 9. In addition to a refreshed user interface for the digital voice assistant Siri and a native News app, the update features a number of new tools specifically tailored for the iPad, Apple's tablet line. One notable change is the iPad's digital QuickType keyboard, which can now switch to a digital trackpad. Using a two-finger swipe, you can select, drag and paste large chunks of text more quickly and easily. Multitasking capabilities have also improved.

We propose Coactive Learning as a model of interaction between a learning system and a human user, where both have the common goal of providing results of maximum utility to the user. Interactions in the Coactive Learning model take the following form: at each step, the system (e.g. search engine) receives a context (e.g. query) and predicts an object (e.g. ranking); the user responds by correcting the system if necessary, providing a slightly improved but not necessarily optimal object as feedback. We argue that such preference feedback can be inferred in large quantity from observable user behavior (e.g., clicks in web search), unlike the optimal feedback required in the expert model or the cardinal valuations required for bandit learning. Despite the relaxed requirements for the feedback, we show that it is possible to adapt many existing online learning algorithms to the coactive framework. In particular, we provide algorithms that achieve square root regret in terms of cardinal utility, even though the learning algorithm never observes cardinal utility values directly. We also provide an algorithm with logarithmic regret in the case of strongly convex loss functions. An extensive empirical study demonstrates the applicability of our model and algorithms on a movie recommendation task, as well as ranking for web search.

This paper investigates methods for quantifying similarity between audio signals, specifically for the task of of cover song detection. We consider an information-theoretic approach, where we compute pairwise measures of predictability between time series. We compare discrete-valued approaches operating on quantised audio features, to continuous-valued approaches. In the discrete case, we propose a method for computing the normalised compression distance, where we account for correlation between time series. In the continuous case, we propose to compute information-based measures of similarity as statistics of the prediction error between time series. We evaluate our methods on two cover song identification tasks using a data set comprised of 300 Jazz standards and using the Million Song Dataset. For both datasets, we observe that continuous-valued approaches outperform discrete-valued approaches. We consider approaches to estimating the normalised compression distance (NCD) based on string compression and prediction, where we observe that our proposed normalised compression distance with alignment (NCDA) improves average performance over NCD, for sequential compression algorithms. Finally, we demonstrate that continuous-valued distances may be combined to improve performance with respect to baseline approaches. Using a large-scale filter-and-refine approach, we demonstrate state-of-the-art performance for cover song identification using the Million Song Dataset.

Using geometric data analysis, our objective is the analysis of narrative, with narrative of emotion being the focus in this work. The following two principles for analysis of emotion inform our work. Firstly, emotion is revealed not as a quality in its own right but rather through interaction. We study the 2-way relationship of Ilsa and Rick in the movie Casablanca, and the 3-way relationship of Emma, Charles and Rodolphe in the novel {\em Madame Bovary}. Secondly, emotion, that is expression of states of mind of subjects, is formed and evolves within the narrative that expresses external events and (personal, social, physical) context. In addition to the analysis methodology with key aspects that are innovative, the input data used is crucial. We use, firstly, dialogue, and secondly, broad and general description that incorporates dialogue. In a follow-on study, we apply our unsupervised narrative mapping to data streams with very low emotional expression. We map the narrative of Twitter streams. Thus we demonstrate map analysis of general narratives.

We present an approach to simultaneously reasoning about a video clip and an entire natural-language sentence. The compositional nature of language is exploited to construct models which represent the meanings of entire sentences composed out of the meanings of the words in those sentences mediated by a grammar that encodes the predicate-argument relations. We demonstrate that these models faithfully represent the meanings of sentences and are sensitive to how the roles played by participants (nouns), their characteristics (adjectives), the actions performed (verbs), the manner of such actions (adverbs), and changing spatial relations between participants (prepositions) affect the meaning of a sentence and how it is grounded in video. We exploit this methodology in three ways. In the first, a video clip along with a sentence are taken as input and the participants in the event described by the sentence are highlighted, even when the clip depicts multiple similar simultaneous events. In the second, a video clip is taken as input without a sentence and a sentence is generated that describes an event in that clip. In the third, a corpus of video clips is paired with sentences which describe some of the events in those clips and the meanings of the words in those sentences are learned. We learn these meanings without needing to specify which attribute of the video clips each word in a given sentence refers to. The learned meaning representations are shown to be intelligible to humans.

There is rising interest in vector-space word embeddings and their use in NLP, especially given recent methods for their fast estimation at very large scale. Nearly all this work, however, assumes a single vector per word type ignoring polysemy and thus jeopardizing their usefulness for downstream tasks. We present an extension to the Skip-gram model that efficiently learns multiple embeddings per word type. It differs from recent related work by jointly performing word sense discrimination and embedding learning, by non-parametrically estimating the number of senses per word type, and by its efficiency and scalability. We present new state-of-the-art results in the word similarity in context task and demonstrate its scalability by training with one machine on a corpus of nearly 1 billion tokens in less than 6 hours.