Historical Trends in Deep Learning
Deep learning has experienced 3 waves of development so far:
From the 1940s to the 1960s, the prototype of deep learning appeared in cybernetics
In the 1980s and 1990s, deep learning manifested as connectionism
It wasn't until 2006 that there was a real revival in the name of deep learning.
Familiar with a few concepts
Automatically learn various algorithms of features and orange types from the data, then the appearance of this model is your rule base.
Where does deep learning fit in artificial intelligence?
Understanding Deep Learning
1. The basic unit of neural network - neuron
In the artificial neuron simulated by the mathematical model, all dendrite signal sources and related strong calculations are processed.
The calculation formula is as follows: s = p1w1+p2w2+p3w3+b
Second, the structure of the neural network
3. The concept of deep learning
A deep neural network (deep learning) is a neural network with at least one hidden layer, i.e. a large number of hidden layers.
The difference between deep learning and traditional methods
Supervised learning
Supervised learning in deep learning includes convolutional neural networks, recurrent neural networks , etc.
unsupervised learning
Unsupervised learning in deep learning includes deterministic autoencoder methods, contrastive divergence methods based on probabilistic restricted Boltzmann machines, etc.
Common methods of deep learning
- autoencoder
- Convolutional Neural Network
- Recurrent Neural Network
Deep Learning Unsupervised Methods Autoencoders
Autoencoders can be used as a feature dimensionality reduction method.
When we use 4 values to represent the four categories:
It is not compact to represent 4 categories with 4 values, and there is the possibility of compressed representation . For example, 2 values can represent these four different numbers.
Deep Learning Supervised Methods Convolutional Neural Networks
Deep Learning Supervised Methods Convolutional Neural Networks
A Supervised Approach to Deep Learning - Recurrent Neural Networks
The source of the recurrent neural network is to characterize the relationship between the current output of a sequence and the previous information . From the network structure, the recurrent neural network will memorize the previous information and use the previous information to affect the output of the following nodes. That is , the nodes between the hidden layers of the recurrent neural network are connected, and the input of the hidden layer includes not only the output of the input layer, but also the output of the hidden layer at the previous moment.
Fu Yuanhui said: "Training in Australia is very hard, I'm already dying, it's just life is better than death." Literally may be angry. "Ghost knows what I've been through, I'm too tired", although the words are hard, but the facial expressions and voice emotions are not, so it is still happy to sum up.
Introducing Reinforcement Chemistry, AIphaGo, and Transfer Learning
reinforcement learning
Don't study, watch TV - parents reprimand, beat
Study hard - reward lollipops
AIphaGo
transfer learning
Various application scenarios of deep learning
security monitor
Smart City
medical health
smart home
Application method of deep learning in intelligent operation and maintenance
The development process of intelligent operation and maintenance
KPI Anomaly Detection Algorithm
Fast clustering of KPIs using autoencoders combined with clustering algorithms
regular pattern
Vibrant mode
unusual pattern
The common KPI data in operation and maintenance is a kind of time series data, which has the characteristics of many data instances and high dimensions. In order to reduce the cost of data analysis work and improve the analysis efficiency, we hope to divide the massive time series data curves into several categories, thereby reducing the number of curves that need to be examined.
Therefore, it is necessary to label large-scale auxiliary KPIs and assist in building a fault propagation chain.
Use LSTM to do KPI trend prediction
write at the end
In recent years, under the background of the rapid development of the AIOps field, the urgent needs of IT tools, platform capabilities, solutions, AI scenarios and available data sets have burst out in various industries. **Based on this, Cloud Wisdom released the AIOps community in August 2021, ** aiming to build an open source banner and build an active user and developer community for customers, users, researchers and developers in various industries. Contribute and solve industry problems and promote technological development in this field.
The community has open sourced the data visualization orchestration platform-FlyFish, the operation and maintenance management platform OMP , the cloud service management platform-Moore platform, Hours algorithm and other products.
Visual Orchestration Platform-FlyFish:
Project introduction: https://www.cloudwise.ai/flyFish.html
Github address: https://github.com/CloudWise-OpenSource/FlyFish
Gitee address: https://gitee.com/CloudWise/fly-fish
Industry case: https://www.bilibili.com/video/BV1z44y1n77Y/
Some large screen cases: