1. Avoids the need to compute the exact Jacobian at each iteration.
2. Suitable for high-dimensional problems where computing the Jacobian is expensive.

1. Convergence is not guaranteed for all problems.

oThe approximation of the Jacobian may become inaccurate over time, requiring periodic reinitialization.