Deep learning study and metamatarials make the invisible

In deep learning study and metamatarials make the invisible visible

By combining purpose-built objects with neural networks, EPFL researchers have shown that sound can be used in high-resolution images.

credit : actu.epfl.ch

Imaging allows us to display an image by remote analysis of the field of light- and sound-wave that you emit or emit. The shorter the wave, the higher its resolution. However, the data level is limited by the wavelength size in question — to date. Researchers at the EPFL Wave Labour have successfully proven that it is long-lasting, and as a result, the wave (in this case a sound wave) can receive information 30 times its length. To achieve this, a team of researchers used a combination of metamatadium - precisely designed elements - and artificial intelligence. Their research, recently published in Physical Review X, creates exciting new opportunities, especially in the field of medical thinking and biyoengineering.

The team’s impressive idea was to bring together different technologies that separated boundaries in the past. One of these is metamaterial: compounds that can, for example, be precisely focused on waves. That said, they are known to have lost their performance by importing illegal signs in a way that makes them difficult to explain. Some artificial intelligence, or direct neural networks can process and quickly process complex information, even though there is a learning curve involved.

Beyond the so-called physics of the limit of distraction, a team of researchers - led by Romain Fleury - did the following experimental work: they first built a library of 64 speakers, each of which could be translated into pixels in the picture. They then used this ton to produce pictures with numerical sounds from zero to nine with precise location details; linear photographic images were taken from a repository of about 70,000 handwritten examples. Across the television the researchers placed a bag containing 39 Helmholtz cosmetics (10 inches and a hole in one end) that formed a metamaterial. The sound produced by this tile was transmitted by metamaterials and captured by four microphones placed a few meters away. The algorithms then locate the sound recorded by the microphone to learn that they can understand and re-insert the first number images.

Also Read : 

• As humans who have read a major area of our brain we can ‘recycle’
  
• Microsoft only prevented Windows 10 users from removing its Google Chrome rival

Beneficial returns

The team achieved almost 90% success rate on their tests. "By producing images with a resolution of only a few inches - using sound waves that are several meters long - we have continued to exceed the limit of distortion," said by Romain Fleury. "Furthermore, the tendency of metamatadium to absorb signals, which were considered to be very complex, turns out to be beneficial when the neural network are really affected. We have found that they work better when there is a greater seizure."

In the field of medical reasoning, using high frequency signals can be very effective. "Long waves mean that the doctors can use very low frequency by the resulting in active acoustic thinking using hard bone tissue. When it comes to the idea of   using electromagnetic waves, long waves do not pose a risk to patients' health. numbers, but rather organic structures, "said Fleury.

• A Python-powered driver Driving program learning tool
• The algorithm created through intensive learning finds potential therapeutic targets across the gene
  
• The biggest AI of 2035 is already here
  
• 5 Free Resources Every Data Scientist Should Start Using Today
  
• The algorithm predicts inverted new material compositions
  
• software developers, you're on the wrong track to machine learning