Do you know MLP (Multilayer Perceptron)? Some people who want to develop something using machine learning may be wondering what exactly MLP is. For those who are interested in AI, it is worth having knowledge of MLP! So this time, I would like to introduce MLP in machine learning in detail!
Table of Contents
 What is MLP (Multilayer Perceptron)?
 3 things MLP can do
 Various types of data can be machinelearned
 Can predict unknown data
 Can solve linearly inseparable problems
 What is the relationship between MLP, machine learning, and deep learning?
 Relationship between MLP and machine learning/deep learning
 What is the difference between MLP and CNN?
 MLPs
 CNN
 What is Python used for machine learning?
 Introduction of neural networks applying MLP
 CNN (Convolutional Neural Network)
 RNN (recurrent neural network)
 LSTM (long short term memory)
 GAN (Generative Adversarial Network)
 How to implement MLP
 What is XOR
 Implementation
 Trends and Future of MLP
 Summary
What is MLP (Multilayer Perceptron)?
MLP (Multilayer Perceptron) is a multilayered perceptron modeled after the human brain. This is the basis of neural networks, one of the machine learning algorithms.
A simple perceptron consists of an input layer and an output layer, but as shown in the figure above, an MLP consists of at least three layers: an input layer, an intermediate layer (hidden layer), and an output layer, and multiple layers are added. It is characterized by
The output of another neuron can be used as the input of another neuron, so computations performed by a simple perceptron can be performed multiple times in a multilayer perceptron at once. Much of today’s machine learning uses extensions to this form of multilayer perceptron.
3 things MLP can do
MLP can do three main things.
 Various types of data can be machinelearned
 Can predict unknown data
 Can solve linearly inseparable problems
Various types of data can be machinelearned
The learning of this MLP uses a method that can machinelearn various data, error backpropagation. This involves setting weights (what data is given importance) to the input data (criteria for judgment), and adjusting the weights so that the error between the output data and the learning data is minimized. . This makes it possible to learn all the weights and classify various data well, so machine learning can be performed on various data.
Can predict unknown data
MLP can infer regression problems. A regression problem is predicting data based on a series of numbers. You can predict other data from the data so far, such as “If you know A, you can predict B.” For example, past sales data can be used to predict next year’s sales.
Can solve linearly inseparable problems
“Linearly inseparable” means that the data distributed on the plane on the Xaxis and Yaxis cannot be classified by a single straight line.
As mentioned earlier, MLP has multiple intermediate layers added, so it is possible to classify multiple classes. A simple perceptron could only process planes linearly, but multilayer perceptrons have the advantage of being able to classify planes into curves. You can see that it is indispensable for machine learning, which requires complex problem processing.
What is the relationship between MLP, machine learning, and deep learning?
So what does MLP have to do with machine learning and deep learning? I will introduce in detail what MLP in machine learning is like.
Relationship between MLP and machine learning/deep learning
Next, I will explain the relationship between MLP and machine learning and deep learning.
In the first place, deep learning derives a solution from a large amount of data. Both deep learning and MLP are one of the machine learning methods that developed the neural network.The artificial neural network imitates the network of the human brain, but as the artificial neural network developed, the structure of MLP was used. Now available.
Originally, only simple perceptrons existed, but MLP was born to fill the shortcomings. MLP has a structure that mimics deep learning.
What is the difference between MLP and CNN?
Similar to MLP is CNN (Convolutional Neural Network). Below, we will introduce the differences between MLP and CNN.
MLPs
MLP is a general neural network in which the input layer, hidden layer, and output layer neurons are all connected. In MLP, data unrelated to output is also passed on to other neurons.
CNN
CNN is also called convolutional neural network. It restricts connections between neurons to only those necessary for output, whereas MLP is fully connected. It also features the ability to share weights.
What is Python used for machine learning?
Python is a programming language used for AI development using machine learning. Python is so popular that it is used by many companies. For example, YouTube and Instagram are famous for using Python. Python is a popular programming language because of its reliability and simplicity of syntax.
We recommend using Python when actually implementing MLP.
Introduction of neural networks applying MLP
Many neural networks that can output data more efficiently by applying MLP have been created. Below are four neural networks.
CNN (Convolutional Neural Network)
A CNN (convolutional neural network) consists of a convolutional layer and a pooling layer inserted between an input layer and an intermediate layer.
RNN (recurrent neural network)
RNN (Recurrent Neural Net) is a neural network with cycles inside. It is used for forecasting timeseries data that cannot be done with MLP. A structure in which output data at one point in time can be used as input for other networks.
LSTM (long short term memory)
LSTM (long shortterm memory) is a type of RNN, and it is characterized by being good at learning time series data.
The shortterm memory of the network can be memorized and utilized for a long period of time. In RNN, the farther apart the data is, the more difficult it is to connect to other related inputs, but in LSTM, it is possible to connect even if the related data is far away.
GAN (Generative Adversarial Network)
A GAN (Generative Adversarial Network) is a network that can learn features from data and generate new data. A GAN consists of a generative network and a discriminative network, and can be learned by making them compete.
By using this, you can generate different image data from multiple image data, or convert an outoffocus image into a clear image.
How to implement MLP
Now let’s actually implement MLP
What is XOR
XOR is also called exclusive OR. It is a logical operation that says OK if only one of the two is correct, and NG if both are correct or both are incorrect.
Simply put

An operation that produces such a result.
Implementation
We will introduce the method of implementation in roughly eight steps.
Step 1: Create a network class
Step 2: Initialize Weights
Step 3: Ingest Dataset
Step 4: Implementing Formal Neurons
Step 5: Implement Forward Propagation
Step 6: Implementing the learning part
Step 7: Implement weight update formulas and determine termination criteria
Step 8: Implement the Test Section
It is good to check the operation one by one while proceeding with these.
Please refer to this article for how to implement
Trends and Future of MLP
As AI using artificial intelligence develops in the future, it is expected that MLP, which is widely applicable, will support more data learning.
MLP is a basic neural network, but many applied neural networks have been developed as introduced. In the future, MLP and MLPapplied neural networks will be used for more complex machine learning.
Summary
How was it. This time, I introduced what MLP can do, its application, and its implementation.
If you are interested in the structure of neural networks such as MLP in machine learning, why not use this as an opportunity to study about neural networks such as MLP?