There are multiple classification methods of Machine Learning based on the emphasis on different aspects.
The Definition of Machine Learning
Machine learning is a multidisciplinary interdisciplinary major, covering probability theory, statistics, approximate theory, and complex algorithms. It uses computers as a tool and is committed to simulating human learning in real-time, and divides the existing content into knowledge structures to effectively improve learning efficiency.
Machine learning has the following definitions:
- Machine learning is a science of artificial intelligence. The main research object of this field is artificial intelligence, especially how to improve the performance of specific algorithms in experiential learning.
- Machine learning is the study of computer algorithms that can be automatically improved through experience.
- Machine learning is the use of data or experience to optimize the performance criteria of computer programs
Classification of machine learning
Over the past few decades, there have been many types of machine learning methods published in research and publications, and there can be multiple classification methods based on the emphasis on different aspects.
Machine Learning Classification based on learning strategy
It is mainly statistical machine learning. Statistical machine learning is based on the analysis of the preliminary understanding of data and the purpose of learning, selecting the appropriate mathematical model, drawing up the hyperparameters, and inputting the sample data, according to a certain strategy, using the appropriate learning algorithm to train the model, and finally using the training. The model analyzes and predicts the data.
Three elements of statistical machine learning:
- Model: Before the model is trained, its possible parameters are multiple or even infinite, so the possible models are also multiple or even infinite, and the set of these models is the hypothesis space.
- Strategy: the criteria for selecting the model with the best parameters from the hypothesis space. The smaller the error (loss function) between the classification or prediction results of the model and the actual situation, the better the model. Then the strategy is to minimize the error.
- Algorithm: the method of selecting a model from the hypothesis space (equivalent to solving the best model parameters). The parameter solving of machine learning is usually transformed into an optimization problem, so the learning algorithm is usually an optimization algorithm, such as the steepest gradient descent method, Newton method, and quasi-Newton method.
Classification based on learning method
- Symbolic induction learning: Typical symbolic induction learning includes example learning and decision tree learning.
- Function induction learning (discovery learning): Typical function induction learning includes neural network learning, example learning, discovery learning, and statistical learning.
- Deductive learning
- Analogical learning: Typical analogical learning includes case (example) learning.
- Analytical learning: Typical analytical learning includes explanatory learning and macro operation learning.
Machine Learning Classification based on learning style
- Supervised learning (learning with a tutor): There are tutor signals in the input data, and the probability function, algebraic function, or artificial neural network is used as the basis function model, and the iterative calculation method is adopted, and the learning result is a function.
- Unsupervised learning (learning without tutor): There is no tutor signal in the input data, the clustering method is used, and the learning result is the category. Typical unsupervised learning includes discovery learning, clustering, and competitive learning.
- Reinforcement learning (enhanced learning): A learning method that takes environmental reversal (reward/punishment signals) as input and is guided by statistics and dynamic programming techniques.
Machine Learning Classification based on the data format
- Structured learning: Structured data is used as input, and numerical calculation or symbolic deduction is used as the method. Typically structured learning includes neural network learning, statistical learning, decision tree learning, and rule learning.
- Unstructured learning: Unstructured data is used as input. Typical unstructured learning includes analogous learning, case learning, explanatory learning, text mining, image mining, and Web mining.
Machine Learning Classification based on learning objectives
- Concept learning: The goal and result of learning are concepts, or learning to obtain concepts. Typical concept learning mainly includes example learning.
- Rule learning: The goal and result of learning are rules, or learning to obtain rules. The typical rule learning mainly includes decision tree learning.
- Function learning: The goal and result of learning are functions, or in other words, to obtain function learning. The typical function learning mainly includes neural network learning.
- Category learning: The goal and result of learning are object categories, or in other words, to obtain category learning. The typical category learning mainly includes cluster analysis.
- Bayesian network learning: The goal and result of learning is a Bayesian network or a kind of learning for obtaining a Bayesian network. It can be divided into structure learning and majority learning.
The application of Machine Learning
- Virtual assistant. Siri, Alexa, Google Now are all virtual assistants. As the name suggests, they will assist in finding information after using voice to give instructions. For answers, the virtual assistant will look up information, recall our related queries, or send commands to other resources (such as phone applications) to collect information. We can even instruct the assistant to perform certain tasks, such as "setting an alarm clock at 7 o'clock" and so on.
- Traffic forecast. We often use GPS navigation services in our lives. When we do this, our current position and speed are saved on a central server for traffic management. These data are then used to construct a mapping of the current traffic. Machine learning can solve the problem of fewer cars equipped with GPS. In this case, machine learning helps to find crowded areas based on estimates.
- Filter spam and malicious software. The email client uses many spam filtering methods. To ensure that these spam filters are constantly updated, they use machine learning techniques. Multi-layer perceptron and decision tree induction are some spam filtering techniques supported by machine learning. More than 325,000 malwares are detected every day, and each code is similar to 90% to 98% of the previous version. System security programs driven by machine learning understand coding patterns. Therefore, they can easily detect 2% to 10% mutations of new malware and provide protection against them.
Machine Learning (ML) in simple terms can be defined as the science of getting computers to act and learn without explicit programming to perform those actions. It has become quite popular in recent years, however, the term itself was coined in 1959 by Arthur Samuel who defined Machine Learning as ‘the field of study that gives computers the ability to learn without being explicitly taught’.
A more recent and formal definition of Machine Learning was created by Tom Mitchell and describes it as a well-defined learning problem – ‘A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E.’
Machine Learning usage has become so common that you probably use it many times a day without even knowing it. For example, when you search on Google or Bing or any other search engine, in the background there is a learning algorithm that has learned how to rank pages based on user queries. Similarly, when you see the photodetection feature on different social media applications or see spam filter filtering out bogus/unwanted emails in your mailbox, behind the scene is a Machine learning algorithm that learns and detect faces or spams emails respectively. A more recent technology use case is the advent of self-driving cars.
This article will focus on the machine learning (ML) pipeline of machine/deep learning infrastructure operations (MLOps).
The following content and technical points will be introduced in this article:
- The definition of the requirements for ML pipelines in the production environment
- The Serverless ML Pipeline solution GitHub based on Alibaba Cloud Serverless Workflow (FnF) and Function Compute (FC)
- The combination guidance of FC and Alibaba Cloud Container Service for Kubernetes (ACK) provided by GitHub. Introductions to task triggering, prediction and inference service deployment, and expose services are also provided.
- Analysis and comparison of this solution and similar solutions. The Serverless ML Pipeline can improve R&D efficiency, reduce O&M costs, and help ML generate value faster.
- Discussion on the selection of the ML infrastructure. FC can complement Kubernetes clusters.
Machine Learning Platform for AI provides end-to-end machine learning services, including data processing, feature engineering, model training, model prediction, and model evaluation. Machine Learning Platform for AI combines all of these services to make AI more accessible than ever.