Machine Learning Courses
Requirements: Basic linear algebra (matrix, vector, matrix vector multiplication), basic probability (probability of random variables and basic attributes), and Calculus
Machine Learning: Course 1
Machine Learning: Course 2
Machine learning algorithms include:
1. Supervised Learning Algorithm
2. unsupervised learning algorithms
3. Reinforcement Learning Algorithm
4. Recommendation System
Machine Learning: Course 3-Supervised Learning
Supervised Learning: training with correct answers
Supervised Learning is the most common technique used to train neural networks and decision trees. These two technologies (Neural Networks and decision trees) are highly dependent on the information provided by a pre-determined classification system. For neural networks, classification systems are used to determine network errors and adjust the network to adapt to them. For decision trees, classification systems are used to determine which attributes provide the most information, in this way, we can use it to solve the classification system problem.
Speech recognition systems using hidden Markov models and Beth networks also rely on some supervisory elements, which are usually used to adjust system parameters to minimize errors in a given input.
In the classification problem,The goal of learning algorithms is to minimize errors in a given input.
If we want to predict the price of a house with an area of 750 square meters, we can best think of fitting these points with a curve and then finding the equation of this curve, then, X is substituted into y to solve the problem. This is supervised learning, because we provide correct results in advance for each piece of data,The above problem is also called a regression problem because the predicted variable Y is continuous.
If the predicted variable is not continuous, but has a category, it is called the classification problem (classification)
Machine Learning: Course 4-Unsupervised learning
Unsupervised learning: uses datasets for training without correct answers (without prior classifications)
Unsupervised learning seems very difficult: instead of telling the computer how to do it, we let it (the computer) learn how to do something on its own. Unsupervised learning generally has two ways.First approachWhen guiding the agent, it does not specify a specific category for it, but uses some kind of incentive system for its success. It should be noted that such training is usually placed in the decision-making framework, because its goal is not to generate a classification system, but to make the most rewarding decision. This idea is a good generalization of the real world. The agent can correct the behaviors.Make incentivesAnd other actionsPenalty.
Some forms of reinforcement learning can often be used for unsupervised learning. On the other hand, it is a very time-consuming learning method for trying to make mistakes. However, this kind of learning may be very powerful because it assumes that there are no pre-classified samples.
Clustering)It is a common unsupervised learning type. The objective of this type of learning is not to maximize the utility function, but to find the approximate points in the training data. Aggregation often finds intuitive classification that matches assumptions.
In an artificial neural network,Self-organizational ing(SOM) andAdaptive Resonance Theory(ART) is the most commonly used unsupervised learning.
Machine Learning notes Parti