Description
Require actual unequal to predicted and return sign(predicted - actual) as the gradient with respect to predicted.
Signed errors preserve direction while absolute or squared errors emphasize magnitude. Their gradients and small-residual approximations represent specific mathematical terms rather than complete performance evaluations.
Require actual unequal to predicted and return sign(predicted - actual) as the gradient with respect to predicted.
When to use this tool
Use Mae Single Observation Gradient when you need an explicit error, gradient, or local loss component.
How to use the tool and how it works
- Enter actual, predicted. Use the units and conventions below.
- Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
- Read gradient. Use the output convention below.
Require actual unequal to predicted and return sign(predicted - actual) as the gradient with respect to predicted.1
| Input field | Meaning and units | Accepted input |
|---|---|---|
| actual | Observed value on the same scale and unit as its prediction. | Required. Finite number. |
| predicted | Predicted value on the same scale and unit as actual. | Required. Finite number. |
| Output field | Interpretation | Returned shape |
|---|---|---|
| gradient | Gradient | Required. Safe integer. |
{
"input": {
"actual": 2,
"predicted": 5
},
"output": {
"gradient": 1
}
}Limitations and assumptions
- Equality is rejected because absolute error is nondifferentiable there; use an explicit equality subgradient policy separately.
- Inputs must satisfy both the field types and the cross-field conditions above. Invalid input causes an error; a successful false flag or documented placeholder is a different result.
- Numeric calculations use binary floating point unless exact decimal-integer arithmetic is explicitly stated. No hidden measurement tolerance or decimal rounding is added; nonfinite or out-of-contract results are rejected. Numeric signed-zero results are exported as zero.
Alternative or Complementary methods
Assess a full error distribution and keep the sign and aggregation conventions consistent across models.
References
Similar or alternative tools
- Mae Subgradient Zero Policy
Return zeroPolicy as the selected absolute-error subgradient at equality, requiring a value in [-1, 1].
- Mse Single Observation Gradient
Compute 2 × weight × (predicted - actual) / denominator, with nonnegative weight and positive denominator.