online gambling singapore online gambling singapore online slot malaysia online slot malaysia mega888 malaysia slot gacor live casino malaysia online betting malaysia mega888 mega888 mega888 mega888 mega888 mega888 mega888 mega888 mega888 Quickly Understand the 6 Types of Neural Networks

摘要: Quickly Understand the 6 Types of Neural Networks



▲來源:Photo by Art Hauntington on Unsplash

The field of artificial intelligence is dominating our life, from chat history to the recommendation and from identification to sorting, there are more and more application scenarios where artificial intelligence involvement is necessary, in artificial intelligence, there is a subfield of Neural Networks which is very popular but many people lack the true understanding of neural networks, people misunderstood the Neural Network with deep learning and interchange the actual meaning of both concepts.

Simple or Feedforward Neural Network

After the introduction of the backpropagation algorithm in 1990 by Geoff Hinton, Neural Network appeared in the developer world, initially Neutral Network only a feedforward that contains a single input layer with multiple hidden layers and a single output layer, no repetition in the execution and process always move forwardly, this model usually suitable for numerical data, due to its design constraints this model never works with sequential data and faces a huge problem when used with image data, the requirement for this model is to manually select the features data and therefore it is difficult for developers to every time manually selects the features from the image data.

Convolutional Neural Network (CNN)

The convolutional neural network (CNN) is better than the feedforward Neural Network because this model always extracts the features from the image data, CNN will effectively retrieve and utilize the neighboring pixel information, it is mostly used for object detection, classification, and detect the bounding objects in image data and very favorable in object detection problem.

Recurrent Neural Network

A convolutional neural network (CNN) is suitable only in classification problem but RNN is the best fit in the classification and meaning of the text data, RNN can help people learn the meaning of text data because RNN worked sequentially over text data and retrieve the meaning of each word because every word depends on the previous word in a sentence.

Long and short-term memory networks (LSTM)

These Neutral Networks are the subcategories of the RNN, they support and increase the temporal memory information by introducing different gates that work like the add or delete button to adjust the network neuron state. These models are helpful in various language modeling tasks and with the stream of input words, this model has effectively done the task of sequential patterns of text data.


This type of Neural Network is basically the attention model network, with the help of attention mechanism this model evaluates the dependencies between input and output, this solution is faster and better in the field of natural language processing (NLP). This model takes each sentence in a paragraph and processes each word sequentially if the paragraph contains many sentences it means the overall execution time also increased. The recent GPT-3 is the largest transformer neural network.

Generative Adversarial Network (GAN)

The GAN usually has two adversarial neural networks, if completely trained and the model understands the nature of the data set, GAN easily produces fake images, and through confrontation and competition, the performance of this Neutral Network will continue to improve over time. GAN is helpful in making video games and special effects moreover realistic textures and unlimited possibilities are possible with GAN based system

Autoencoder or Self-encoder

This type of Neutral network is approximately creating the mapping of input to output, it first compresses the data features into low dimensional data set then reconstructs the output from that dataset. This model is helpful in anomaly detection.

轉貼自Source: medium

若喜歡本文,請關注我們的臉書 Please Like our Facebook Page:    Big Data In Finance




  • 找不到回應