Chapter OneArtificial IntelligenceIntroduction to Machine Learning and Deep Learning
1.1 AI: Weak AI and Strong AI
machine learning
Use an algorithm to generate a model after training with a large amount of data, and use this model to achieve the prediction effect
is a branch of artificial intelligence supervised learning unsupervised learning reinforcement learning
deep learning
Mimic the way human neural networks work
is a branch of machine learning multilayer perceptron deep neural network recurrent neural network
The reasons for the accelerated development of artificial intelligence in recent years
1 Big data distributed storage and computing
2GPU TPU parallel computing
CPU contains several cores optimized for blood sucking processing
GPUs can have up to thousands of small and efficient cores that can harness the power of parallel computing
Deep learning simulates the way neurons work with a large number of matrix operations. The characteristic of matrix operations is that a single operation is very simple, but requires a large number of operations. It is especially suitable for parallel computing using GPUs for parallel computing through a large number of cores.
1.2 Introduction to Machine Learning
Consists of features and labels
features: data features such as temperature, wind direction, wind speed, season, air pressure
label: The predicted target such as the weather (rainy, sunny, foggy, etc.) or the specific value of the temperature
Two stages: training predictions
Training: The training data is obtained by feature extraction to obtain features and labels, which are then put into the algorithm for training to obtain the model for prediction
Prediction: new data -> feature extraction to get features and put them into the model to get prediction results
1.3 Machine Learning Classification