A parameter we “just give” to algorithms, not learn via training. Hyperparameters are evaluated by cross-validation sets and fine-tuned hyperparameters are your ultimate goals. For instance, distance function selection is a hyperparameter setting.
It does not always produce the same output for a given input. A few components of systems that can be stochastic in nature include stochastic inputs, random time-delays, noisy(modeled as random) disturbances.