We introduce a Maximum Entropy model able to capture the statistics of melodies in music. The model can be used to generate new melodies that emulate the style of the musical corpus which was used to train it. Instead of using the $n-$body interactions of $(n-1)-$order Markov models, traditionally used in automatic music generation, we use a $k-$nearest neighbour model with pairwise interactions only. In that way, we keep the number of parameters low and avoid over-fitting problems typical of Markov models. We show that long-range musical phrases don't need to be explicitly enforced using high-order Markov interactions, but can instead emerge from multiple, competing, pairwise interactions. We validate our Maximum Entropy model by contrasting how much the generated sequences capture the style of the original corpus without plagiarizing it. To this end we use a data-compression approach to discriminate the levels of borrowing and innovation featured by the artificial sequences. The results show that our modelling scheme outperforms both fixed-order and variable-order Markov models. This shows that, despite being based only on pairwise interactions, this Maximum Entropy scheme opens the possibility to generate musically sensible alterations of the original phrases, providing a way to generate innovation.

The great gathering of genre fans in the city that never sleeps is officially over. New York Comic-Con has closed its doors after a weekend of surprise comic book cameos, never-before-seen TV premieres and geeky celebrations. But the pictures of the cosplay at NYCC will live on forever, because the East Coast convention goers gave great costume this year. Here's a collection of a few of our favorite costumes, should you dare to be inspired this Halloween season. Jessica Jones in the scarcely cosplayed version of her (now retired) alter superhero character Jewel, Luke Cage in Power Man attire, and Marvel's brand new Iron Man character Riri Williams.

As I explored the data, I became particularly interested in the question of overfitting. You can follow the links at the top of the page to see how the same language model reacts to sentences from a different corpus. This post has some more background and discusses a few interesting examples.

It's one that Alexa told me last week while I was reviewing the new Echo Dot. That's the thing -- however intelligent your AI might be, that doesn't mean it's got a good sense of humor. To really set itself apart from the likes of Alexa and Siri, the next great artificially intelligent assistant is going to need some better material. For its part, Google is looking to step up its new Assistant's game with writers who have honed their skills with gigs at some beloved places All of this is part of a bid to move the products from simple Q&A'bots to the kind of thing a user might actually form an honest to goodness emotional connection with, as though Her weren't enough of a cautionary tale already. According The Wall Street Journal's vaguely ominously titled article Your Next Friend Could Be a Robot, Google Home has already picked up some folks from places like The Onion and Pixar, somewhat echoing Anki's recent hiring spree for its adorable upcoming robot, Cozmo.

As Google Home gets closer to launch, the AI assistant is not only getting smarter, but also a little more friendly and -- hopefully -- a whole lot funnier. As the Wall Street Journal's Christopher Mims notes in a piece about friendly AI like Siri, Alexa and Google Assistant, Alphabet's big play for the space includes hiring up comedy writing alumni of Pixar and The Onion. While none of the major AI assistants on the market today or coming soon are truly "artificial intelligence" (that is: the device itself doesn't actually understand the conversation) people have a natural tendency to form an emotional connection with the little robot voice in the kitchen speaker. So, subtle improvements like a witty joke or unexpected bit of humor can go a long way to improving the user experience, especially as voice and conversation becomes the interface itself. Although Siri and Alexa already have a few jokes in their repertoire, they tend to be pretty bad and definitely don't have quite the same punch as the topical humor of "America's Finest News Source."

This is a place to share machine learning research papers, journals, and articles that you're reading this week. If it relates to what you're researching, by all means elaborate and give us your insight, otherwise it could just be an interesting paper you've read. Please try to provide some insight from your understanding and please don't post things which are present in wiki. Preferably you should link the arxiv page (not the PDF, you can easily access the PDF from the summary page but not the other way around) or any other pertinent links.

It seems like there isn't a week in which deep learning doesn't come up as achieving some kind of remarkable task. I understand that one of the powers of deep learning is that it is capable of learning the features. This capacity seems totally decoupled from the underlaying problem. So basically I read this as "no matter what problem you have... You can use deep learning".

The film studio 20th Century Fox has called in IBM Watson, the supercomputer, to create the trailer for its upcoming AI horror/thriller, Morgan. IBM Watson has done many things; beat human contestants in quiz shows, created bespoke recipes and described the contents of photos. Now it has become the first-ever AI to produce a film trailer for the new sci-fi film. IBM researchers fed Watson more than 100 horror film trailers cut into separate moments and scenes. It performed a series of visual, sound and composition analyses on each scene to get an idea of how to create the dynamics of a trailer.

Robot" creator Sam Esmail has teased what fans could expect of Joey Bada's character Leon in Season 3 after they were left surprised by the post-credits scene in the Season 2 finale of the critically-acclaimed USA Network TV series. "Leon is definitely a character to look out for in in season_3.0," Bada, who, in addition to being an actor, is also a successful rapper and founder of hip-hop collective Pro Era, talked about what his character will be up to, but he was careful not to go into detail. "We'll just have to wait and find out. It's looking good for Leon, though; really looking good for him," the rapper told MTV when asked if the post-credits scene in the Season 2 finale means he will be getting a "killer" plotline in "Mr.

This is one egg you don't want to break. Branded as the "first smart assistant designed especially for the kitchen," this cute little gadget is a voice-operated culinary helper that boasts both a display and personality. Brought to market by Internet of Things-focused company RnD64, HelloEgg wants to put a whole team of cooking helpers in your kitchen by way of one electronic device. The egg-shaped piece of technology depends upon both artificial intelligence and a group of real humans to help you cook and improve your kitchen skills. With step-by-step cooking videos and the ability to ask specific questions when a person on a screen just won't cut it, HelloEgg might just be how you wean yourself off Seamless.