AI Techniques and Tools Through Python. Supervised Learning: Classification Methods, Ensemble Learning and Neural Networks

FIRST CONCEPTS IN ARTIFICIAL INTELLIGENCE

1.1 

ARTIFICIAL INTELLIGENCE

Today, the amount of data generated by both humans and machines far exceeds the ability of humans to absorb, interpret and make complex decisions based on that data. Artificial intelligence is the foundation of all machine learning and the future of all complex decision-making processes. Artificial Intelligence combines mathematical algorithms and Machine Learning, Deep Learning and Big Data techniques to extract the knowledge contained in data and present it in a comprehensible and automatic way.

Artificial intelligence (AI) is a branch of computer science that seeks to develop systems capable of performing tasks that traditionally require human intelligence, such as pattern recognition, decision-making, natural language processing and learning. Within this field, machine learning techniques play a crucial role, especially in recent developments in AI.

Among the areas of application of Artificial Intelligence are the following:

Machine Learning: Algorithms that allow machines to learn from data and improve over time without being explicitly programmed.

Natural Language Processing (NLP): Allows machines to understand, interpret and generate human language. An example is ChatGPT.

Computer Vision: Enables machines to see, identify and process visual objects and scenes.

Robotics: Creation of robots that can interact with the environment and perform physical tasks.

Artificial Neural Networks: Models inspired by the human brain for complex tasks such as speech, image and pattern recognition.

1.2 

ARTIFICIAL INTELLIGENCE TECHNIQUES

Artificial intelligence techniques are the methods and approaches used to develop systems that can perform tasks that normally require human intelligence. These techniques fall into several categories, each with its own set of approaches and algorithms. The following paragraphs describe the main AI techniques

1.2.1  Machine Learning (ML)

Machine learning is a branch of AI that allows machines to learn from data without being explicitly programmed to perform specific tasks. There are several approaches within machine learning:

Supervised learning: The model learns from labelled data, i.e. data with a known response. The goal is for the model to predict the correct output for new data based on what it has learned. Examples include linear regression, classification and neural networks.

Unsupervised learning: The model finds patterns or structures in unlabelled data. It is used for tasks such as customer segmentation, clustering analysis and dimensionality reduction. Examples include the K-means algorithm and Principal Component Analysis (PCA).

Reinforcement learning: The model learns to make decisions through trial and error, receiving rewards or penalties based on the actions it takes. It is common in control tasks and games, such as learning autonomous agents. Examples include Q-learning and Deep Q Networks (DQN).

1.2.2  Artificial Neural Networks (ANN)

Neural networks are models inspired by the human brain, consisting of layers of neurons that process information. They are used for complex tasks such as speech recognition, image recognition and natural language processing. Neural networks are generally classified as follows:

Deep Learning: A subcategory of machine learning that uses neural networks with many layers, known as deep neural networks. They are especially effective for complex pattern recognition in large amounts of data. Examples include convolutional networks (CNNs) for computer vision and recurrent networks (RNNs) for sequence processing, such as in machine translation.

Generative Antagonistic Networks (GANs): Use two competing neural networks: one generates false data (such as images) and the other evaluates it. They are popular in the creation of content such as images and music.

1.2.3  Evolutionary Algorithms

Evolutionary algorithms are methods inspired by biological evolution, such as natural selection and mutation. These algorithms seek to find optimal solutions to complex problems by simulating evolutionary processes.

Genetic Algorithms (GA): They use mechanisms such as selection, crossover and mutation to evolve solutions to problems. They are used in the optimisation of processes and the search for solutions in large and complex spaces.

Evolutionary programming and genetic programming: Variants of genetic algorithms that are used for the evolution of computer programs or algorithms.

1.2.4  Fuzzy Logic

Fuzzy logic is a mathematical approach that handles uncertainty and imprecision, allowing systems to make decisions based on information that is not entirely accurate. It uses values between 0 and 1 (instead of just 0 and 1, as in classical logic) to represent degrees of truth.

It is applied in control systems, such as smart thermostats and robots that make decisions with uncertain information.

1.2.5  Natural Language Processing (NLP)

Natural language processing (NLP) is concerned with enabling machines to understand and generate human language. Techniques in NLP include:

Sentiment analysis: Determining whether a text expresses a positive, negative or neutral opinion.

Language models: Such as Transformers (e.g. GPT-3), which are deep learning models capable of generating coherent text, translating languages, summarising texts, etc.

Named Entity Recognition (NER): Identifying names of people, places, dates, etc. in a text.

1.2.6  6. Expert Systems

Expert systems are AI programs that simulate the knowledge and skills of an expert in a specific domain. These systems use knowledge bases and inference rules to make decisions or solve problems in areas such as medical diagnosis, finance and industrial maintenance.

1.2.7  Optimisation Algorithms

Optimisation algorithms search for the best (or close to the best) solution to a given problem, usually by maximising or minimising an objective function.

1.2.8  Computer Vision

Computer vision is a field that allows machines to see and understand the visual world. It uses various machine learning techniques and neural networks to process images and videos. Some applications include:

Object recognition: Identify and classify objects in pictures or videos.

Image segmentation: Dividing an image into meaningful regions or segments.

Facial recognition: Identification of persons from their facial features.

1.2.9  Automatic Reasoning

Automatic reasoning allows machines to make logical decisions based on rules. It involves the application of formal logic to deduce new facts from existing facts.

Rule-based systems: They use a set of logical rules to solve problems or make inferences.

1.2.10  Intelligent Agents

Intelligent agents are autonomous systems that perceive their environment, make decisions and act on it. Agents can be simple, such as a virtual assistant, or more complex, such as autonomous robots. They use learning and reasoning techniques to interact with their environment and accomplish goals.

Artificial Intelligence techniques are often combined and applied to real-world problems to create intelligent systems that can learn, adapt and make decisions autonomously.

1.3 

MACHINE LEARNING

Machine learning techniques teach computers to do what comes naturally to humans: learn from experience. Machine learning algorithms use computational methods to extract information directly from data. The algorithms improve their performance adaptively as the number of samples available for learning increases. Machine learning uses two types of techniques: supervised learning, which trains a model with known input and output data to predict future results, and unsupervised learning, which finds hidden patterns or intrinsic structures in the input data.

1.3.1  ​Descriptive or unsupervised learning techniques

In these techniques there are no dependent or independent variables, so they are called interdependence techniques or descriptive techniques. Unsupervised learning finds hidden patterns or intrinsic structures in data. It is used to draw inferences from datasets consisting of input data without labelled responses. Segmentation is the most common unsupervised learning technique. It is used to find hidden patterns or clusters in data. Applications for clustering include gene sequence analysis, market research, preference analysis, etc.

The most commonly used descriptive or unsupervised learning techniques in data analysis are the following:

  Exploratory data analysis

  Dimension reduction

  Principal component analysis

  Factor analysis

  Simple correspondence analysis

  Multiple Correspondence Analysis

  Multidimensional scaling

  Correlation