In data science and engineering, a surrogate (often called a surrogate model) is a simplified model that approximates a much more complex, expensive, or slow process. Instead of repeatedly running a heavy simulation or real-world experiment, we train a surrogate to mimic the original system. Once trained, this lighter model can produce predictions in milliseconds.
Think of it as a “stand-in brain”: it learns patterns from historical data and then answers new questions quickly.
Surrogates are widely used in fields such as machine learning, computational engineering, finance, and scientific research—anywhere direct evaluation costs too much time or money.
