Feature Extraction
The process of transforming raw data into a set of meaningful numerical features that machine learning models can effectively use for prediction.
Traditional vs Deep Learning
Traditional ML requires manual feature engineering (domain expertise to create useful features). Deep learning automatically learns features from raw data through its layers. Transfer learning uses pre-trained models as feature extractors.
Common Techniques
TF-IDF for text, HOG/SIFT for images (traditional), embeddings from pre-trained models (modern), PCA for dimensionality reduction, and domain-specific features (technical indicators for finance, molecular descriptors for drug discovery).