Shulou(Shulou.com)06/01 Report--

How to set the number of neurons and network layers in the big data neural network, I believe many inexperienced people are helpless about this, for this reason this article summarizes the causes and solutions of the problem, through this article I hope you can solve this problem.

lead

For beginners in machine learning, there are two questions that are confusing: 1) Given a neural network, how do you determine the number of middle layers? 2) How to determine the number of neurons in each middle layer?

An important theorem of neural network model is the principle of universal approximation, which means that there exists a large enough network to achieve any desired accuracy. However, this theorem does not give a specific method for calculating the number of layers and neurons in the network.

The only way to truly understand which architecture works best and most violently is to try all of them and choose the best one. For neural networks, however, this is very difficult because each model takes a lot of time to train. One approach is to train a model that is too large and then prune the model by eliminating weights that do not contribute much to the network.

In fact, there is no universal way to determine the number of layers and neurons in a network. But we can approach the problem indirectly in two somewhat abstract ways.

01

Deeper network models

A deeper network model means that the model has more middle layers (hidden layers). Its role is to allow the model to compute more complex features. For example, in a convolutional neural network, the first few layers of the model represent "low-level" features, such as edges; the last few layers represent "high-level" features, such as faces, body parts, etc. For unstructured data (such as images), deeper models fully exploit the high-level complexity of the data.

02

Broader network model

More layers of the model (deeper) means more complex features can be created, and a wider model (more neurons in a single layer) means more features can be created. Some problems only require extracting simple features from the data, but many of these simple features are needed, so that a broader model can be useful. In general, the dimensionality to the end of the network becomes narrower and narrower, because complex features carry more information than simple features, the end of the network needs to reflect the depth of the model more than the width.

After reading the above, do you know how to set the number of neurons and network layers in a big data neural network? If you still want to learn more skills or want to know more related content, welcome to pay attention to the industry information channel, thank you for reading!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.