Lei feng’s network (search for “Lei feng’s network” public attention) by writer Tian yuandong, Facebook researcher at the artificial intelligence laboratory.
We pace the field quickly, because the content of the paper backward, all conference papers, two or three a year, half-year research cycle.
But because it is a double-blind review, even if there are others doing similar, submission when the two do not know, draft also issued in the deal at the same time. Emergence of Arxiv, allows anyone to put their work published on the Web, allows peers to see, competition in this area even further. Routine a week or even every day to brush Arxiv look good article, more importantly see what works and they currently do overlap. Whenever there is a new idea and new idea, had to get it, as long as a month or two is no longer you, because someone else had already beat us to it, directly on article hammered you killed, also called out. Computer go had bloody show before this, everything happens in a few months, we let the other players of the peers were depressed, and then AlphaGo let us depressed, yesterday’s zhongxingpengyue, and today, the less people know, although inconstancy, it was understandable. Ted Baker Galaxy S4 Case
Program framework and open source are exacerbating this process, reproduce someone else’s pain is not only time consuming, but now is the use, the cost is zero. Want to stay ahead, hide their tips tricks that are meaningless, eateries advertising an article published in a year, who are not open source or later source, who would have been forgotten in the corner of history. NIPS last year everyone was talking about ResNet, ResNet out within 24 hours of our group finished the program and has already begun training in ImageNet, and soon confirmed that it works. TensorFlow open source, the backbone of the Group on the busy 3:30 A.M. to complete the evaluation. Generally say they work hard but do not publish proceedings, unless substantial hardware resources like the AlphaGo system, otherwise someone will judge, and fast. CIFAR was once an Arxiv report better results and immediately was challenged in reddit, and then open the code, training on the set was found in tests of low-level errors, had to sadly withdrawn drafts.
Chung is meeting tastes have also changed. The intent of the meeting was to enjoy the latest advances in, and go to a meeting now is nostalgia, full of old article was half a year ago, have a sense of which go back in and inexplicably comic. We often talk about not the lecture or poster as far as the eye can reach, but just come in Arxiv’s latest, “just”, probably refers to the last week, yesterday, may also refer to a few hours ago, had just as little and complacent, twinkling like a bucket of ice water head under water, spaced out, awake at night, not knowing the taste. Compared to the other half-year review cycle, iteration faster update speed is unprecedented. I sometimes wonder, except woke up hungry fat wakes up after dreaming awake, also add Arxiv woke up, dreams of gods is done by others, or dream of posting a few days ago encountered stern warning under sweat suddenly woke up, and then wiping my face, thank goodness.
Under such pressure, people’s way of thinking has changed. So far, we have not to essence of deep understanding of deep learning, many research ideas are simple and crude, in today’s computing resource abundance, for the quality of their subjective judgment, far less often than write code machine is fast and accurate. Many thoughts right or wrong, no matter how ridiculous it looked ridiculous, try again, often faster than thought, and faster than the brain. Daily cell phone similar with us, this shift in thinking is good or bad, there is no way to judge. Such practices can be classified as impetuous and eager for quick success and his ilk, runs counter to the traditional academic approach-but compared to the words, such schemes continue to produce results, promotes the progress of technology, to make a better system. We can complain about Google machine scheduling too much too fast, the other side is also preparing a large number of GPU’s ambitious battle on the ground. In any case, can be expected to be, by a person a piece of paper and a pen to overcome challenges, admirable little efficiency; and let the computers involved in reasoning and judgment, and will eventually become a required course for all men are created.
From this trend Shang view, can fast using various tool reached target of, can fast reading and the wrote code of, has powerful engineering capacity of, in employment market Shang are has huge advantage; if again has based solid, can fast read understand articles, from gets new knowledge and analogy, experience rich, see paper can caught points pain points start, from numerous idea in the trim thought, that is veteran has; again Shang a layer, has strategy vision, can from thousands of articles in the see trend vision, clever set problem, to minimum cost gets maximum returns, That is a first-class masters, can stand on the front line, directly promoting developments in the field. However to do this step, you need to stop and accumulated experience in rapid iteration, a long time before they can take effect.
Artificial intelligence behind the popular, is the number of days you come to me to fight, in the high dimensional space siege pulling village, hand-to-hand. Although tired, but it is an appeal and hope in this field. And more frightening than bitter and tired, is comfortable when it was dismissed for skills obsolete, hovering between heaven and Earth, there’s no place to go to. I can’t imagine that kind of despair, so it was close behind the pace of the times, at least, we can see the track technological development, for it is a force. In this field, as long as the tool used correctly, yet artfully, everyone has potential as a superhero, a good algorithm framework can serve thousands of people, and the 24-hour work, do not make mistakes, not fatigue, which had completely unthinkable. The saying goes:
Juan stream at Ling handicraft skills, surging to this vast expanse;
However, silica Sands a handful of soil, better week lang million soldiers.
Lei feng’s network: the first known by the author column in this paper: the far East anecdotes, please contact empower and retain the full information, articles must not be modified.
In addition, in August this year, Lei feng’s network in Shenzhen will hold an unprecedented global influence of artificial intelligence and robotics Innovation Conference. When Lei feng’s network (search for “Lei feng’s network”, public interest) will be published “artificial intelligence and robot Top25 list of innovative enterprises”. At present, we are visiting related companies in the area of artificial intelligence, robots, sift through them eventually elected to the list list of companies. If you want to join our list, please contact: 2020@leiphone.com. Ted Baker Galaxy S4 Case
Best iPhone Cases from China 