Machine Learning Fundamentals

Machine learning enables computers to learn from data without explicit programming, driving advancements across various fields through algorithms like supervised and unsupervised learning.

Core Principles

• Machine learning (ML) algorithms learn patterns from data to make predictions or decisions.
• Supervised learning uses labeled data to train models for classification and regression tasks.
• Unsupervised learning works with unlabeled data to discover hidden structures and patterns through methods like clustering and dimensionality reduction.
• Deep learning, a subset of ML, utilizes neural networks with multiple layers to process complex data.
• The machine learning pipeline involves data collection, preprocessing, model training, evaluation, and deployment.
• Mathematical concepts like statistics, linear algebra, and calculus are foundational to understanding and implementing ML algorithms.

Formulas

• $ \text{Accuracy} = \frac{\text{Vrais Positifs} + \text{Vrais Négatifs}}{\text{Total}} $
• $ \text{Recall} = \frac{\text{Vrais Positifs}}{\text{Vrais Positifs} + \text{Faux Négatifs}} $
• $ \text{Precision} = \frac{\text{Vrais Positifs}}{\text{Vrais Positifs} + \text{Faux Positifs}} $
• $ \text{F1 score} = 2 * \frac{(\text{Recall} * \text{Precision})}{(\text{Recall} + \text{Precision})} $
• $ \text{Normalisation} = \frac{X - X_{min}}{X_{max} - X_{min}} $
• $ \text{Standardisation} = \frac{X - \mu}{\sigma} $
• $ \text{Gini}(T) = 1 - \sum_{j=1}^{m} (\frac{|T_j|}{|T|})^2 $
• $ \text{Gain de Gini}(X, T) = \text{Gini}(T) - \sum_{i=1}^{k} \frac{|T_i|}{|T|} \text{Gini}(T_i) $
• $ d(X, Y) = \sqrt{\sum_{i=1}^{n} (x_i - y_i)^2} $
• $ J(\theta) = \frac{1}{2m} \sum_{i=1}^{m} [h(x_i) - y_i]^2 $
• $ \nabla J(\theta) = \frac{1}{m} X^T (X\theta - y) $
• $ \theta = (X^T X)^{-1} X^T y $
• $ \theta^{(k+1)} = \theta^{(k)} - \alpha \nabla J(\theta^{(k)}) $

Pitfalls to Avoid

• Overfitting: Model performs well on training data but poorly on new data.
• Underfitting: Model is too simple and cannot capture underlying patterns.
• Data Leakage: Information from outside the training data inadvertently influences the model.
• Bias in Data: Skewed datasets can lead to biased model predictions.
• Choosing the Wrong Metric: Evaluating model performance with inappropriate metrics.