What does it mean for you to do "y * dy/dx" and "y (dy/dx)2"? I mean, are you doing elementwise operations? Also could you report the magnitude on the plot of both "y" and "dy/dx"? My guess is that the norm of the gradient might be very small and thus y dy/dx 2 and you would basically be doing SGD. Also, I don't think you finetuned well SGD RMSprop and so on as they should also give excellent results on MNIST.

"We should study Science Fiction in order to understand what someday could become Science Fact." This is the time of year when everyone makes his or her predictions for 2018. I have my predictions as well, but wanted to do something a bit more fun. So I thought I'd look backwards to the state of technology 50 years ago to gain some insights that we can use to make projections about 2018. That is, what "predictions" made in the 1950's might tell us about 2018.

As an artificial intelligence researcher, I often come across the idea that many people are afraid of what AI might bring. It's perhaps unsurprising, given both history and the entertainment industry, that we might be afraid of a cybernetic takeover that forces us to live locked away, "Matrix"-like, as some sort of human battery. And yet it is hard for me to look up from the evolutionary computer models I use to develop AI, to think about how the innocent virtual creatures on my screen might become the monsters of the future. Might I become "the destroyer of worlds," as Oppenheimer lamented after spearheading the construction of the first nuclear bomb? I would take the fame, I suppose, but perhaps the critics are right.

With intelligent automation marching steadily toward broader adoption, media coverage of this historic technology disruption is turning increasingly alarmist. "New study: Artificial intelligence is coming for your jobs, millennials,"1 announced one business news outlet recently. "US workers face higher risk of being replaced by robots,"2 declared another. These dire headlines may deliver impressive click stats, but they don't consider a much more hopeful--and likely--scenario: In the near future, human workers and machines will work together seamlessly, each complementing the other's efforts in a single loop of productivity. And, in turn, HR organizations will begin developing new strategies and tools for recruiting, managing, and training a hybrid human-machine workforce. Notwithstanding sky-is-falling predictions, robotics, cognitive, and artificial intelligence (AI) will probably not displace most human workers. Yes, these tools offer opportunities to automate some repetitive low-level tasks. Perhaps more importantly, intelligent automation solutions may be able to augment human performance by automating certain parts of a task, thus freeing individuals to focus on more "human" aspects that require empathic problem-solving abilities, social skills, and emotional intelligence. For example, if retail banking transactions were automated, bank tellers would be able to spend more time interacting with and advising customers--and selling products.

Just wondering if there are some typical optimization problems or specific challenges that are fundamental to machine learning, i.e. something many or most frameworks should be capable of solving? For example, I'd like to try optimizing the coefficients of an ANN, and would love to have some basic, fundamental input/output problems to solve.

Ever since God created man, the Almighty has assured us of our sovereignty over the planet and control over all of its creatures. It is for this reason we fear the idea of aliens who may be more intellectually advanced than us or the emergence of intelligent machines that might think independently. Our fear of the appearance of those who surpass our intelligence is evident from the science fiction and futuristic films that speak about this subject (such as Terminator or Star Wars films). However, this fear of the emergence of entities that may surpass our intelligence has not stopped people from continuing to make machines that think for themselves. In America and Japan in particular, research is rapidly advancing towards this goal.

LG has just revealed that it will be showcasing its artificial intelligence solution for its upcoming high-end TVs at CES 2018 next week. The South Korean electronics company is also showing off its advanced image processor at the Las Vegas exhibition. On Wednesday, LG took to its online newsroom to announce its ThinQ AI and its new image processor for its upgraded OLED and SUPER UHD TVs. The main goal behind the first invention is to make it easier for consumers to interact and remotely control their televisions. On the other hand, the latter is designed to ensure that the picture quality and viewing experience of the new television screens will be unrivalled in the industry.

Many, especially science fiction writers, play somewhere in the middle. They write about technologies that don't yet exist. Or they assume the possibilities of technologies far ahead of their time. Melbourne, AUSTRALIA: Kellie Shaw (R) inspects one of Leonardo da Vinci's most famous designs which is recognised as the ancestor to the modern helicopter and is part of an exhibition featuring 50 models of 15th century inventions by Da Vinci, in Melbourne 04 July 2006. This craft is made of linen, reeds and iron thread and would have been operated by four men rotating a shaft. The exhibition focuses on four themes: mechanical, military, hydraulic and flying machines with each model built according to Da Vinci's drawing and are crafted from materials available in 15th century Italy.

The woman in the photo seems familiar. She looks like Jennifer Aniston, the "Friends" actress, or Selena Gomez, the child star turned pop singer. She appears to be a celebrity, one of the beautiful people photographed outside a movie premiere or an awards show. She was created by a machine. The image is one of the faux celebrity photos generated by software under development at Nvidia, the big-name computer chip maker that is investing heavily in research involving artificial intelligence. At a lab in Finland, a small team of Nvidia researchers recently built a system that can analyze thousands of (real) celebrity snapshots, recognize common patterns, and create new images that look much the same -- but are still a little different.