M
M
mIka012021-07-04 15:37:18
Neural networks
mIka01, 2021-07-04 15:37:18

Questions about the neural network?

Hello, I have a number of questions about neural networks. Programming for me is something like a hobby, so if the questions are stupid, then don't get angry.

What is the difference between multilayer neural networks and multilayer perceptron?
How does unsupervised learning work?
How does self-classification of a neural network work? (this article tells about the fact that the neural network, without learning, can, as it were, independently classify according to the same criteria)
How does genetic algorithm learning work? (neural networks with additions of topologies)

Is it possible to represent a neural network in a matrix form? That is, not through an algorithm, but through linear mathematics.

When a video is shown on YouTube on how a neural network learns to park, I have a question. If learning takes place in one cycle, that is, there is a task and then the answer, how does the neural network understand at what stage it made a mistake? How does training happen on such videos (conditionally)?

A convolutional neural network was created to classify images, but are there scanning neural networks that can reproduce an image? If so, how does it work, because there is a convolution layer in a convolutional network?
The convolution layer in a convolutional neural network is designed to reduce the image, then if the image itself is not large, you can use the standard network?
Is it possible to replace the convolution layers in a convolutional neural network with other image reduction algorithms (for example, Seam carving), and then apply finding the lines as in the human eye and feed this data to the network?

How to determine the required number of neurocity layers and how many neurons should be in each word?

Thanks in advance for your replies.

Answer the question

In order to leave comments, you need to log in

1 answer(s)
D
dmshar, 2021-07-04
@dmshar

Why so few questions? They would say right away - "come on, give me a course of neural networks here for a quickie."
But I wonder, since there are so many questions, is it really so difficult to open any book on machine learning, in which there are already answers to all your questions, to read and understand? And even for another three dozen that you will definitely have when you find the answers to those that you have already asked. What's stopping you? And when you don’t find answers in books, then you should look for help from the forum.
Yes, and to draw information from popular science videos for younger students, and then ask on the forum "how does it work" is also not necessary. Answers - look in those same books, they are definitely there.
PS By the way, although programming is related to neural networks, these are completely different areas of knowledge. Therefore, the fact that you have programming in you is "something like a hobby" - well, in general, it does not add any useful information to the question.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question