I just want to emphasise for any readers that the density ratio approximation is tight only when the discriminator is optimal. Given that they update the discriminator even less than the generator they are essentially optimising a quite loose approximation to a likely loose bound on the data log likelihood.

Camera phones become more sophisticated every year, and it's changing the landscape of photography. Professionals are using their smartphones in place of hefty DSLRs, and indie filmmakers are shooting everything from music videos to feature-length dramas. With advanced features and tons of free editing software, the camera phone may become the lens of the future. Tip in a Tip: Apple's camera app is good, but there are better free apps to capture and share your memories. That little camera is useful for so much more than snapshots and home videos.

Since the plagues of the Old Testament, we have contemplated the Apocalypse, the world rising in vengeance as men, women and children scurry across the brutal landscape of a lost paradise. Our doomsday stories and how they scroll and flash before us have changed since the parchment days of the Bible. But we remain fascinated by the specter of our demise, whether the end is wrought by deities, our own folly or imposed by outside forces like monsters, asteroids and aliens that have haunted us since Orson Welles' 1938 "War of the Worlds" radio broadcast. Few of our dystopias, however, are as frightening as the planet gone asunder, polluted and destroyed by humanity's amorality, recklessness and greed. Film and literature -- to say nothing of our private insecurities -- resound with a world that freezes, boils, chokes, cracks with earthquakes, dwindles with resources and succumbs to pestilence and disease.

Students of marketing are worried. They fear that they will already have been replaced by the time they graduate and look for career positions. And some of this fear is warranted, of course. AI has become the new "darling" of marketing, as an ever-increasing number of uses keep being discovered. But, the picture is not as bleak as marketing students may be painting. What marketers need to understand is that AI can relieve them of all of the mundane manual tasks, provide them with amazing insights, and allow them to spend time on the creative aspects of their jobs.

Students of marketing are worried. They fear that they will already have been replaced by the time they graduate and look for career positions. And some of this fear is warranted, of course. AI has become the new "darling" of marketing, as an ever-increasing number of uses keep being discovered. But, the picture is not as bleak as marketing students may be painting. What marketers need to understand is that AI can relieve them of all of the mundane manual tasks, provide them with amazing insights, and allow them to spend time on the creative aspects of their jobs.

We have witnessed the first wave of artificial intelligence (AI) in journalism in the form of chatbots, automated story generation, and machine learning techniques applied to news. The big tech companies have pushed AI to the center of their product strategies. How far along is the news business in incorporating these tools into the newsroom, and understanding the broad implications for journalism? At this event, we will examine how real examples of AI have been used in news, and what we can expect in the near future. We will discuss the importance of this growing field and identify priorities for data journalists, researchers, investigative reporters, and newsrooms.

After going through YOLO / Darknet ( C!), Caffe ( Works only commit a8b0c1a), MXNet (Not yet stable) detectors, its great to see Google providing state of the art pre-trained models for variety of architectures ranging from MobileNet to R-FCN along with reliable/reproducible training scripts. Though I now feel bad about spending time to get YOLO/Keras to train instead of just waiting.

Why did they not use an (additive) attention mechanism for the encoder and decoder? It's heavily proven to improve gradient flow. Perhaps I'm missing something big here that would make this training procedure not work.

Here's some news that's not fake: Not everything you can read on the internet is true. Trouble is, it can be hard to know truths from untruths, and there's evidence untruths travel faster. Many hands have been wrung in recent months over what to do about made-up news stories created to convert social media shares into page views, ad dollars, and perhaps even political traction. The modest first results from an effort to crowdsource machine learning technology to help stem the flood of falsity are a reminder that machines may help us grapple with fake news--but only if humans take the lead. Late last year, Facebook's director of AI research Yann LeCun told journalists that machine learning technology that could squash fake news "either exists or can be developed."

Toymaker Sphero is releasing an interactive toy in the form of a Spider-Man smart speaker. Unlike the company's previous toys, this interactive Spider-Man toy is stationary and controlled through voice commands. Sphero is mostly known for creating the BB-8 toy from "Star Wars" or the Lightning McQueen toy from the Pixel film "Cars." But unlike those two, Sphero's Spider-Man toy focuses more on interacting with kids through actually speaking to them. What sets Sphero's Spider-Man apart is that it is capable of guiding a child through comic book storylines.