Real Number
Any number on the continuous number line, including irrationals.
- Uncountable; computers use float approximations.
- Most ML weights, gradients, losses are reals.
- 0.1 + 0.2 ≠ 0.3 in floating point.
When to use: Any continuous quantity: weights, gradients, probabilities, learning rates.
Example: Learning rate 0.001, weight −1.4327, softmax prob 0.6843 - all reals stored as floats.