Hello, I would like to know examples of applications, and in general, where and how neural networks and AI are used in applications. I myself write on java for android, and I would like to know if it is realistic to combine these classes: "Study neural networks and AI and practice and implement them in Mobile Applications?

  • AI is not only neural networks. What tasks do you want to solve? - Alexander Muksimov

2 answers 2

As my practice has shown, it is possible, but writing a neural network in the application itself is not the best idea. When I started to study neural networks, I wrote my neural network in the form of a library for android with 0. I use it in 2 of my projects: stock market analysis and analysis of film preferences. In both cases, everything works, but it gives a very large load on the CPU of the phone and it literally becomes a stick on which you can fry an egg. If the network is not complete, there is not much data to process, but after reading a bunch of articles I realized that it is much better to release the neural network separately on the server and communicate with it over the network. In my opinion, the best framework for NA is TensorFlow from Google.

    It is quite possible to combine. Neural networks are used in Prisma for processing photos, and in Snapchat for masking (real-time facial recognition using a neural network is used, and after that effects are applied). In the mobile application Auto.ru, neural networks make it possible by photo to find the right vehicle for sale. In general, the use of neural networks is wide, so some tasks in mobile applications can be solved very effectively with their help. Especially those associated with pattern recognition.

    An example implementation of neural networks in Java

    The neural network is trained on the back-propagation error algorithm

    neural network type-multilayer perceptron

    • But these neural networks, which are also written in java in Mobile Applications? - Martinez Toni
    • Usually in large applications, data (for example, a photo) of a user is sent to the server, and it is already being processed using neural networks. The implementation, in terms of language selection, is different for everyone. Most often, python is used for such solutions, and for it there are already many ready-made libraries for neural networks. But the neural network can be implemented in any language, including Java. It all depends on the task and resources. - Clarence
    • In addition, your application does not have to interact with the server, the neural network can be written in the application itself. - Clarence