The neural network in a smartphone: which way it can help users?

NEUROSEED
NEUROSEED
Aug 27, 2017 · 4 min read

Last year the world has seen many developments which had been created on an Artificial intelligence (AI) using neural networks. One of the recent development was the airframe developed by engineers from Microsoft. For its handling a special algorithm was developed. And there are many such examples.

The majority of these developments are created for a certain type of media: aircraft, robots, computers, etc. However, the accelerating pace of world development requires more mobility, which makes smartphones the leaders in popularity among all the existing gadgets. Therefore, it is important to know and to penetrate how a neural network can be integrated into mobile phones.

Neural network: what is it and what is its purpose

The neural network is the direction of researches conducted in the branch of artificial intelligence, which is based on attempting artificial reproduction with the help of special algorithms of the nervous system of a living organism. The main feature of any neural network is the ability to constantly learn and correct previous errors.

A neural network is used to solve problems of varying complexity, requiring significant analytical calculations. Often with the help of this technology the following tasks are solving:

-Classification. The network performs the distribution of data according to the specified parameters;

-Prognostic. The algorithm provides the ability to predict the next step;

-Recognition. This is the most commonly used function. Implies the recognition of persons or objects at the images.

Today it performs only these three tasks, but in the future it may expand its functionality to a higher level. At this stage of the technology development developers have managed to increase the speed of the particular computing elements — the “neurons”. This helped to reduce the time spent on the network “training”.

Features of the algorithm realization on the mobile phone

Neural network and similar to them algorithms on portable devices (smartphones and tablets) often are implemented in not adaptive look (for example, Viola-Jones’s method which automatically takes away red eyes on a photo). Adaptive realizations are observed much more less. And their adaptation happens not for any task, but for the certain behavior and features of the person. An example of such realization are paramedical tasks (e.g. tracking the state of health, activity etc.).

Today there is an opportunity in the Metatrader for the mobile device (for iOS or Android) to carry out manual programming of any neuronet for the purpose of its further training in forecasting (e.g. on rates of currency pairs). As a result the trade neural network robot having narrow specialization appears. That’s why it’s a long way to the universality here.

“Weight” of the neuronets

The complexity of integration of a neuronet on portable devices is in energy consumption and taking much time. That’s why neural networks often work at cloud servers. In turn servers obtain data from smartphones or table devices then send results of the carried-out analysis.

Vivienne Se together with the colleagues developed a new highly energy efficient computer chip which is optimized for neuronets. It allows the AI powerful systems to function locally on portable devices.

Also experts have developed the new analytical method capable to define how much energy the neural network will consume on a certain type of the hardware. For lowering the amount of the consumed energy on mobile devices the method of “energy saving cropping” is used.

Despite the existence of certain approaches of the neuronets usage on smartphones, there is no optimal solution for such combination.

Neuronet on the mobile phone — its present opportunities

Recently there appear mobile applications for smartphones which are implemented on mathematical algorithms of the neural network. The most known of them are :

  • Prisma. The application which will transform photos into pictures of the world famous artists. The program spends for photo processing only 1 second. It can be used not only in the entertaining purposes, but also in computer graphics, animation, etc. Today the algorithm has filters which imitate styles of 21 artists;

-Mlvch. Prisma analog, it differs from the competitor by the more difficult study. As a result finally the detailed image comes out, as the appendix has 2,5 times more filters;

-Clarifai. This application is capable to recognize the image on video;

-Findface. Allows to find by means of only one photo person’s profile in a social network of VKontakte;

-Авто.ру service. Gives the chance to define model and a make of the car by a photo. The algorithm works by the principle of “computer sight”;

-Ostagram. Allows to carry out an overlay of styles and also to redraw pictures;

-Deepomatic. The application makes the analysis of images which were distributed by the user in the Internet, in conclusion it offers goods presented in online store.

Many applications which are developed for smartphones and function on neural network, perform one and the same operation with images. For example recently Google has got the Belarusian startup AIMatter which carries out processing of visual images on smartphones. The application can not only change hair color on photos, but also improve a background and also carry out other visual effects.

Besides applications with various arsenal of visual tools, algorithms of neural network on the mobile device can perform other functions. For example the startup Ava has developed the application watching user’s health. It makes the analysis of photo with food or dishes and issues data on existence in them of concrete types of nutritional supplements and also counts calories.

The neural network working on the portable device (e.g.the mobile phone, the tablet) allows the user to edit images generally. However it is one of the functions available to a neuronet. Therefore introduction of this technology on smartphones haven’t exhausted all gadget qualities. Niсhу hasn’t been engaged yet and waits for new innovative solutions.

NeuroSeed just offers a new format of the decision for portable devices which is based on combination of the two most popular technologies of nowadays — AI and a blockchain. The project is aimed on creation of “pocket AI” which is available to all in the form of a mobile application. The new digital product which NeuroSeed plans to create will allow to make the smartphone more universal and responding the modern world.

)
Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade