Online Learning

machine-learning

Definition

Online Learning

Online learning is a paradigm where the model is updated incrementally as instances arrive sequentially from a continuous data stream. Formally, at each step $t$, the learner receives $x_t$, predicts ${\hat{y}}_t$, receives the true feedback $y_t$, and immediately updates its parameter vector ${\theta }_{t-1}\to {\theta }_t$ based on the loss $L({\hat{y}}_t,y_t)$.

Adaptability and Efficiency

Incremental Updates (Concept Drift): Online learning is highly suitable for non-stationary environments where the underlying distribution $P(X,Y)$ changes over time, allowing the model to adapt continuously to new patterns.

Memory Efficiency: Unlike batch learning, online methods do not require storage of the full dataset, as they only maintain the current model state and process individual samples or mini-batches. However, they can be more sensitive to noise in individual data points.