In recent years, several major players such as Google, Microsoft and Facebook have created their own AI R&D team and achieved some remarkable results.
On November 9, 2015, Google announced TensorFlow open source, a huge database for fast gradient machine learning on the GPU. Some articles speculate that TensorFlow will bring about an artificial intelligence revolution, saying that Google's move is bold, because Torch (maintained by Ronan Collobert of Facebook's Artificial Intelligence Lab) has provided similar deep learning open resources, while Professor Yoshua Bengio The lab has been long-term maintenance development of Theano (a pioneer in deep learning, a revolutionary software for the general public). In an article by Wired, Cade Metz described TensorFlow as Google's "artificial intelligence engine."
This article is about open source databases for linear algebra and derivation calculations, and even the title is exaggerated. In many other news reports, I was amazed at Google’s code as a public resource. From a more technical side, from exaggerated praise to pouring cold water, there are various reactions. Soumith Chintala released a set of standards for all competing software packages, providing a quantitative assessment that shows that the first version of TensorFlow lags behind Torch and Caffe, especially in convolutional neural networks.
The neural network uses hardware and software to build a neural network similar to the human brain, which dates back to the 1980s, but it was not until 2012 that Krizhevsky and Hinton began to invent neural networks on GPUs. technology. GPUs were originally dedicated processing chips for games and other high-performance graphics software, but they have proven to be very suitable for driving neural networks.
Google, Facebook, Twitter, Microsoft and many other companies now use GPU-driven artificial intelligence to handle a variety of tasks, including image recognition and security applications. Krizhevsky and Hinton later joined Google.
A research team at Microsoft designed a neural network that is much more complex than a "typical design" that can perform as many as 152 layers of complex mathematical operations, while typical designs typically have only six to seven layers. This indicates that in the next few years, companies such as Microsoft will be able to use a large cluster of GPUs and other dedicated chips to greatly enhance a wide range of artificial intelligence services, including image recognition, including recognizing speech and even understanding human natural expression. Speaking. But building such a large neural network is extremely difficult.
In order to determine the working mode of each layer and how it communicates with other layers, different specific algorithms need to be deployed to each layer, but this is an extremely difficult task. But Microsoft also has skills here. They designed a computing system that would help them build these networks. Researchers can identify a number of large neural network deployments that may be useful, and then the computing system can cycle through a series of possibilities until the best choice is determined.
According to Adam Gibson, chief researcher at deep learning startup Skymind, similar practices are now more common. This is called "hyper parameter opTImizaTIon".
He said: "People can let a group of machines run, run 10 models at a time, and then find the best one. They can input some basic parameters (based on intuition), and then the machine determines what is based on this. "The best solution." Gibson said that a company acquired by Twitter last year, Whetlab, provided a similar "optimal" approach to neural networks.
It is expected that 2016 will be a watershed in machine emotion recognition, and emotions will become a powerful new channel for us to interact with machines, and due to the development of camera technology and computer vision algorithms, future machines will pass our human facial expressions and eye movements. The ability to understand our body, body language, way of speaking, and even raising our heads will be greatly enhanced.
Fernando De la Torre of the Carnegie Mellon University Robotics Institute invented a particularly powerful facial recognition software called IntraFace. His team used machine learning to teach IntraFace how to recognize and track facial expressions in a way that works with most faces. Then they created a personalized algorithm that allowed the software to analyze the emotional expression of the individual. Not only accurate but also efficient, the software can even run on mobile phones.
In the future, machines will understand our emotions more and our interaction with machines will become more abundant. JusTIne Cassell of Carnegie Mellon University studied the application of virtual peers in the education industry. She found that when virtual companions can respond appropriately to the emotional state of students, even when they laugh at them on some occasions, students will participate more actively. Will learn better. It's not hard to imagine how much the business world would like to use this feature. Advertisers, marketers, and filmmakers can get more specific information about the customer base.
In terms of the combination of medical and AI, the current doctor's consultation is mainly based on the medical image information left by the patient's current examination, and the impact of past medical history, family history and test results is almost ignored in the diagnosis. But imagine that if the patient's physical data can be recorded in real time, continuously, and there is a smart enough medical diagnostic system to compare this data with data from patients with similar symptoms worldwide. Based on the current clinical medical research and guidance, and comprehensive diagnosis, is it accurate and scientific?
A biosensor research company called Sentrian has developed a medical system that can do the above. The company is headquartered in Florida, USA, and is committed to research related to machine learning. The smart medical system is now in clinical testing. They want to create a medical system that allows doctors to focus on patient body data in real time to make better, earlier, and more personalized diagnostics.
Wireless biosensors can now be used to collect simple or more complex body information such as body temperature, heart rate, blood oxygen saturation and blood potassium levels. Usually, each remote patient wears only one or two sensors at a time, and their data can be directly analyzed by a doctor. If the patient continues to wear multiple sensors, the resulting data will be very large.
Sentrian's medical system uses machine learning algorithms for analysis after collecting patient data. The system contains information on changes in body data for chronic diseases (including heart disease, diabetes, chronic obstructive pulmonary disease, etc.), patient information will be matched against this information, and the system will be diagnosed early by observing subtle associations. Information such as heart rate, blood pressure, and oxygen saturation will also be transmitted to the cloud for analysis and notified to the doctor if necessary.
Fiber optic lighting is total different from any other regular lighting we see. Fashion, creative, elegant and graceful. The whole lighting soft glowing effect will catch most of attention and stand out from the crowd.
Light become flexible and could go through an unexpected directly, not only go in straight line.
Right now, fiber optic lighting apply to indoor and outdoor situation. Not only the traditional Chandelier, star ceiling and curtain but also outdoor landscape decoration. For holiday, wedding, party & events lighting idea, fiber optic lighting is great choice.
Safety use is another great point, people could tough it freely as it is total insulation. And it is totally waterproof, very popular in public use.
Lighting illumination project could find good product in Fiber Optic Light as G-Lights provides customized service also.
Fiber Optic Light
Side Glow Fiber Optic,Fiber Optic Lamp,Fiber Optic Decoration,Fiber Optic Pool Light
ZHONGSHAN G-LIGHTS LIGHTING CO., LTD. , https://www.glightsled.com