In recent years, the rise of Artificial Intelligence and the widespread use of Machine Learning have revolutionized the way we tackle complex real-world challenges. However, due to the diverse nature of data involved, choosing the right algorithm is crucial to achieve efficient and effective solutions. Therefore, understanding the strengths and weaknesses behind different Machine Learning algorithms, and knowing how to adapt them to meet specific challenges, can become a fulcral skill to develop.
Furthermore, since the choice of algorithm greatly depends on the specific task and data involved, it's clear that there is no "Master Algorithm" (No algorithm can solve every problem). For example, while Linear Discriminants effectively delineate boundaries in data that is linearly separable, they struggle to capture relationships in more complex, higher-dimensional spaces.
This Project focuses on the following topic:
With no Master Algorithm, is it possible to improve a existing Machine Learning Algorithm in characteristics it struggles the most?
Therefore, after choosing a Machine Learning Algorithm and gaining a thorough understanding of its theoretical and empirical aspects, we aim to refine it, specifically targeting its weaknesses in solving classification problems.
Nowadays, since singular Machine Learning Algorithms can fall short to predict the whole data given, we decided to study an Ensemble Algorithm. Since these Algorithms can combine outputs of multiple models it makes them more prone to better address more complex problems and provide better solutions.
Consequently, after careful consideration, we decided to focus on enhancing the AdaBoost Algorithm M1, which is employed in binary classification problems.
|
AdaBoost (Adaptive Boosting) is a type of ensemble learning technique used in machine learning to solve both classification and regression problems. It consists on training a series of weak classifiers on the dataset. Therefore, with each iteration, the algorithm increases the focus on data points that were previously predicted incorrectly.
|
|
As a result, the AdaBoost algorithm builds a model by considering all the individual weak classifiers which are weighted based on their performance. Consequently, classifiers with higher predictive accuracy contribute more to the final decision which reduces the influence of less accurate ones in the final prediction.
- Authors → Gonçalo Esteves and Nuno Gomes
- Course → Machine Learning I [CC2008]
- University → Faculty of Sciences, University of Porto
README.md by Gonçalo Esteves