Microsoft researchers in depth on the development of learning technology, once again beyond their own fruit. Asia Research Institute staff published an article about deep learning of academic papers, and said Microsoft's latest deep learning system on an index beyond human beings.
The paper said deep learning of the latest Microsoft systems distinguish between 2012 ImageNet image database, error rate is only 4.94%, compared to the average person's 5.1%. The task of identifying images are generally asked participants to identify objects in the image, then this image categorized in the correct category. Microsoft researchers in an academic paper says, in this Visual recognition tasks, such a result (recognition error rates) beyond human levels for the first time.
https://www.youtube.com/watch?v=MfzZimP6ygQ
The academic paper published not only attracted the attention of stakeholders in the field of artificial intelligence, meanwhile, proved the strength of in depth learning technology in Microsoft. Of course, can also cause Facebook and Google's talent competition. Deep learning is a new field of machine learning research, analyse their motivation is to build, simulate the human brain neural network learning, which allow the computer to mimic the human brain thinking mechanism to explain the data, such as images, sounds and text.
Microsoft researchers also said that deep learning system through promotion of the latest Google GoogLeNet system performance, probably by 26%, in order to achieve beyond the capacity of humanity. Original GoogLeNet system image resolution error rate of 6.66%.
Funny thing is, the Microsoft researchers do not believe that computer image resolution is better than humans. They said:
Although in this particular database, our achievements that result, but this does not mean that on the image recognition, computer images are better than humans. Basic category in recognition of the time, such as daily necessities, or common sense, the computer there are obvious errors, humans could not. However, we still believe that our results suggest that on the image recognition algorithm has a huge potential in the computer reaches human levels.
via:venturebeat
1087 votes
Millet router Hello Kitty cases
As millet router upgrade, new millet router global launch Broadcom Broadcom 4709C dual-core 1.4GHz CPU, with separate 512MB system Flash memory with 4 Gigabit wired network interface, built-in Seagate/Toshiba professional supervisory-level hard drive, normal 1TB, optional 6TB, support for mobile phones and cameras automatically upload.
View details of the voting >> Hello Kitty iphone 6 plus
No comments:
Post a Comment