Machine Learning Series

Part 6: ML Algorithms

A lot of things happen in the vast universe of Machine Learning (ML), and algorithms help us find our way. They're not just math or code; they're what make machine learning work. They give data life and turn boring numbers into deep insights and new ideas.

?? Why Algorithms Matter: At the heart of every ML solution, there's an algorithm. It's the algorithm that decides the trajectory of your data, transforming it into a powerful tool for prediction, classification, and much more. It can be hard to keep track of all the different algorithm names when you are supposed to know what they all mean and where they fit in.

According to Jason Brownlee he want to give you two ways to think about and categorize the algorithms you may come across in the field.

• The first is a grouping of algorithms by their learning style.
• The second is a grouping of algorithms by their similarity in form or function (like grouping similar animals together).

Algorithms Grouped by Learning Style

Various methods exist for algorithms to analyze a problem by interacting with the experience or environment, or whatever we choose to label the input data. Exploring learning styles that algorithms can adopt is a common practice in machine learning and artificial intelligence textbooks. Exploring the primary learning styles or models that algorithms can adopt, along with specific examples of algorithms and compatible problem types. Organizing machine learning algorithms in this taxonomy is beneficial as it prompts consideration of the input data and model preparation process to choose the most suitable one for optimal results.

Check Part 3 and Part 4.

Part 3: https://www.dhirubhai.net/pulse/machine-learning-series-yash-negi-9krmc/

Part 4: https://www.dhirubhai.net/pulse/machine-learning-series-yash-negi-ibdpe/

Semi-Supervised Learning:

The data consists of both labeled and unlabeled examples.There is a specific prediction challenge where the model needs to understand the data structures and make accurate predictions.Typical issues include classification and regression.Algorithms provided as extensions to other flexible methods assume how to model the unlabeled data.

Algorithms Grouped By Similarity

Algorithms are commonly categorized based on their similarities in terms of function. For instance, methods based on trees and methods inspired by neural networks. Some algorithms can fit into multiple categories, such as Learning Vector Quantization, which is inspired by neural networks and is an instance-based method. There are also categories that share the same name to describe the problem and the class of algorithm, like Regression and Clustering.We could address these cases by either listing algorithms twice or by choosing the group that is the most suitable subjectively. I prefer this latter method of avoiding duplicating algorithms to maintain simplicity.

Regression Algorithms

Regression involves modeling the connection between variables, which is continuously improved by assessing the accuracy of the model's predictions.Regression methods are commonly used in statistics and have also been incorporated into statistical modeling. This might be perplexing as regression can be used to describe both the type of problem and the type of algorithm. Indeed, regression is a systematic procedure.

The most popular regression?algorithms are:

• Ordinary Least Squares Regression (OLSR)
• Linear Regression
• Logistic Regression
• Stepwise Regression
• Multivariate Adaptive Regression Splines (MARS)
• Locally Estimated Scatterplot Smoothing (LOESS)

Instance-based?Algorithms

This learning model involves a decision problem with instances or examples of training data that are considered crucial to the model.These methods usually create a database of example data and then compare new data to the database using a similarity measure to determine the best match and make predictions. Due to this, instance-based methods are also known as winner-take-all methods and memory-based learning. The emphasis is on the representation of the stored instances and the similarity measures applied between instances.

The most popular instance-based algorithms are:

• k-Nearest Neighbor (kNN)
• Learning Vector Quantization (LVQ)
• Self-Organizing Map (SOM)
• Locally Weighted Learning (LWL)
• Support Vector Machines (SVM)

Regularization?Algorithms

An extension developed for another method, usually regression methods, that penalizes models according to their complexity, preferring simpler models that are more effective at generalizing.

The most popular regularization algorithms are:

• Ridge Regression
• Least Absolute Shrinkage and Selection Operator (LASSO)
• Elastic Net
• Least-Angle Regression (LARS)

Decision Tree?Algorithms

Constructing a model of decisions based on actual attribute values in the data.Decisions branch out in tree structures until a final prediction is made for a specific record. Trained on data for classification and regression problems are decision trees. Decision trees are known for their speed and accuracy, making them a popular choice in the field.

The most popular decision tree algorithms are:

• Classification and Regression Tree (CART)
• Iterative Dichotomiser 3 (ID3)
• C4.5 and C5.0 (different versions of a powerful approach)
• Chi-squared Automatic Interaction Detection (CHAID)
• Decision Stump
• M5
• Conditional?Decision Trees

Clustering?Algorithms

Clustering, like regression, describes the class of problem and the class of methods.

The most popular clustering algorithms are:

• k-Means
• k-Medians
• Expectation Maximisation (EM)
• Hierarchical Clustering

Bayesian Algorithms

Bayesian methods are those that?explicitly apply Bayes’ Theorem for problems such as classification and regression.

The most popular Bayesian algorithms are:

• Naive Bayes
• Gaussian Naive Bayes
• Multinomial Naive Bayes
• Averaged One-Dependence Estimators (AODE)
• Bayesian Belief Network (BBN)
• Bayesian Network (BN)

Association Rule Learning?Algorithms

Extracting rules that explain relationships between variables in data is a common practice in machine learning.This process can uncover valuable and practical connections within extensive sets of data that can benefit a company.

The most popular association rule learning algorithms are:

• Apriori algorithm
• Eclat algorithm

Artificial Neural Network Algorithms

Artificial Neural Networks are models inspired by the structure and/or function of biological neural networks.These are a type of pattern matching commonly utilized for regression and classification tasks, encompassing a vast subfield with numerous algorithms and variations for various types of problems.

The most popular artificial neural network algorithms are:

• Perceptron
• Multilayer Perceptrons (MLP)
• Back-Propagation
• Stochastic Gradient Descent
• Hopfield Network
• Radial Basis Function Network (RBFN)

Deep Learning Algorithms

Deep Learning methods represent a contemporary advancement in Artificial Neural Networks that take advantage of ample affordable computation.They focus on constructing significantly larger and more intricate neural networks. As mentioned earlier, numerous techniques revolve around extensive datasets of labeled analog data, including images, text, audio, and video.

The most popular deep learning algorithms are:

• Convolutional Neural Network (CNN)
• Recurrent Neural Networks (RNNs)
• Long Short-Term Memory Networks (LSTMs)
• Stacked Auto-Encoders
• Deep Boltzmann Machine (DBM)
• Deep Belief Networks (DBN)

Other Machine Learning Algorithms

• Computational?intelligence (evolutionary algorithms, etc.)
• Computer Vision (CV)
• Natural?Language Processing (NLP)
• Recommender Systems
• Reinforcement Learning
• Graphical Models
• more......