Feature extraction has gained increasing attention in the field of machine learning, as in order to detect patterns, extract information, or predict future observations from big data, the urge of informative features is crucial. The process of extracting features is highly linked to dimensionality reduction as it implies the transformation of the data from a sparse high-dimensional space, to higher level meaningful abstractions. This dissertation employs Neural Networks for distributed paragraph representations, and Latent Dirichlet Allocation to capture higher level features of paragraph vectors. Although Neural Networks for distributed paragraph representations are considered the state of the art for extracting paragraph vectors, we show that a quick topic analysis model such as Latent Dirichlet Allocation can provide meaningful features too. We evaluate the two methods on the CMU Movie Summary Corpus, a collection of 25,203 movie plot summaries extracted from Wikipedia. Finally, for both approaches, we use K-Nearest Neighbors to discover similar movies, and plot the projected representations using T-Distributed Stochastic Neighbor Embedding to depict the context similarities. These similarities, expressed as movie distances, can be used for movies recommendation. The recommended movies of this approach are compared with the recommended movies from IMDB, which use a collaborative filtering recommendation approach, to show that our two models could constitute either an alternative or a supplementary recommendation approach.

This paper presents a novel latent variable recurrent neural network architecture for jointly modeling sequences of words and (possibly latent) discourse relations between adjacent sentences. A recurrent neural network generates individual words, thus reaping the benefits of discriminatively-trained vector representations. The discourse relations are represented with a latent variable, which can be predicted or marginalized, depending on the task. The resulting model can therefore employ a training objective that includes not only discourse relation classification, but also word prediction. As a result, it outperforms state-of- the-art alternatives for two tasks: implicit discourse relation classification in the Penn Discourse Treebank, and dialog act classification in the Switchboard corpus. Furthermore, by marginalizing over latent discourse relations at test time, we obtain a discourse informed language model, which improves over a strong LSTM baseline.

Similarity metrics are a core component of many information retrieval and machine learning systems. In this work we propose a method capable of learning a similarity metric from data equipped with a binary relation. By considering only the similarity constraints, and initially ignoring the features, we are able to learn target vectors for each instance using one of several appropriately designed loss functions. A regression model can then be constructed that maps novel feature vectors to the same target vector space, resulting in a feature extractor that computes vectors for which a predefined metric is a meaningful measure of similarity. We present results on both multiclass and multi-label classification datasets that demonstrate considerably faster convergence, as well as higher accuracy on the majority of the intrinsic evaluation tasks and all extrinsic evaluation tasks.

Researchers often summarize their work in the form of posters. Posters provide a coherent and efficient way to convey core ideas from scientific papers. Generating a good scientific poster, however, is a complex and time consuming cognitive task, since such posters need to be readable, informative, and visually aesthetic. In this paper, for the first time, we study the challenging problem of learning to generate posters from scientific papers. To this end, a data-driven framework, that utilizes graphical models, is proposed. Specifically, given content to display, the key elements of a good poster, including panel layout and attributes of each panel, are learned and inferred from data. Then, given inferred layout and attributes, composition of graphical elements within each panel is synthesized. To learn and validate our model, we collect and make public a Poster-Paper dataset, which consists of scientific papers and corresponding posters with exhaustively labelled panels and attributes. Qualitative and quantitative results indicate the effectiveness of our approach.

I have 34 years as an engineer here at Intel. Almost all but about the last, I don't know, four or five has been mainly on the manufacturing side; all of our silicon manufacturing. Which, in many ways makes you a maker because you're producing a million chips a day. When I look at a 3D printer, I look at it as not only what can I build with it, but I understand exactly how that machine works. I could take it apart and put it back together.

Whether we realize it or not, machine learning is already a part of our everyday lives. Think about a simple Google search, a quick query to Apple's Siri, an afternoon visit to Facebook, or, of course, a great product recommendation on Amazon. As the prevalence of high-performance computing continues to grow, the thought of computers doing things we could only dream of is exhilarating and, for many, almost magical. This is especially true when it comes to marketing. Yet, as we apply machine learning to marketing and unique business goals, oftentimes all of that mystery and intrigue turn to disillusion and struggle with the practical applications.

A U.S. government-backed committee has reportedly recommended that the Federal Aviation Administration loosen restrictions on the use of commercial drones, a crucial first step to wider business applications for unmanned aerial vehicles. The FAA's Aviation Rulemaking Committee of industry leaders and agency officials late Friday recommended easing restrictions that prohibit businesses from flying unmanned aerial vehicles over populated areas in some cases. FAA spokesman Les Dorr says the agency received the report on Saturday and needs time to review the recommendations. The committee's report, obtained by the Associated Press, recommends the creation of four categories of drones. Devices weighing half a pound or less would have essentially no limits on flying over crowds if the manufacturer could ensure a 99 percent chance that it would not injure someone even if it fell on them.

Bengaluru: Infosys Foundation, the philanthropic arm of the IT services company, said it will give a Rs.24-crore grant over the next three years to Delhi-based Indraprastha Institute of Information Technology (IIIT) to set up a research centre for artificial intelligence (AI). For Infosys, which carries out its Corporate Social Responsibility through the foundation, supporting educational institutions has been a fundamental part of its CSR strategy. The grant comes at a time when Infosys chief executive Vishal Sikka, who has a doctorate degree in artificial intelligence from Stanford University, has been trying to increasingly build out capabilities around the strategic focal points of automation and artificial intelligence in the company. The proposed Infosys Center for Artificial Intelligence will facilitate work on both fundamental and applied aspects of AI and focus on areas such as robotics, machine learning, computer vision, AI for software systems, large-scale data analytics. The research will draw on real-time data to develop a deeper understanding of AI for social benefits, and the application of AI in education and related areas, said the release.

Microsoft CEO Satya Nadella has reshaped the company since taking over two years ago. Windows is still important, but it's no longer the only platform that matters: Microsoft is releasing software and supporting app development for Apple's iOS, Google's Android, and even its old enemy Linux. The infighting and aggressive dismissal of competitors is mostly gone. And Nadella has embraced cloud computing -- the idea that some customers don't want to run their own technology but would prefer to outsource it -- turning Microsoft into the clear No. 2 in the category after Amazon. We caught up with Nadella fresh off the company's Build conference for developers last week in San Francisco and ahead of the new Envision conference for business leaders, which kicks off Monday in New Orleans. Matt Rosoff: There was a lot of talk last week at Build about chatbots and artificial agents and "conversation as a platform." That idea is not new, right? I think I heard Bill Gates talking about it 15 years ago.

Schools are on the front lines in coping with attention deficit hyperactivity disorder (ADHD). More school-age kids are getting diagnosed with it each year (more than one in 10, according to the most recent National Survey of Children's Health) and the classroom is where kids often have their biggest problems with impulse control and an inability to sit still and focus. Some kids take medicine to control these symptoms, but many do not. And so principals and teachers are tremendously interested in non-medical therapies they can use at school to help children. Fortunately, it's an exciting time in ADHD research, thanks to developments in neuroscience, and psychologists hope they will find new tools for schools.