Description
Apply max(probability, floor) with both inputs in (0, 1].
Negative log probability scores an observed event. Flooring probabilities changes that score and support rules distinguish zero contribution from an infinite mismatch.
Apply max(probability, floor) with both inputs in (0, 1].
When to use this tool
Use Information: Log Loss Probability Floor Bias when you need one explicit log-probability policy or support diagnostic.
How to use the tool and how it works
- Enter probability, floor. Use the units and conventions below.
- Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
- Read result. Expand result to inspect the returned fields.
Apply max(probability, floor) with both inputs in (0, 1]. Return original and floored negative natural logarithms and bias equal to floored loss minus original loss.1
| Input field | Meaning and units | Accepted input |
|---|---|---|
| probability | Probability in the stated interval; zero handling follows the calculation above. | Required. Finite number. |
| floor | Floor | Required. Finite number. |
| Output field | Interpretation | Returned shape |
|---|---|---|
| result | Result | Required. Object. |
| result.appliedProbability | Applied Probability | Calculated result member: number. |
| result.originalLoss | Original Loss | Calculated result member: number. |
| result.flooredLoss | Floored Loss | Calculated result member: number. |
| result.bias | Floored loss minus original loss in nats; nonpositive. | Calculated result member: number. |
{
"input": {
"probability": 0.01,
"floor": 0.1
},
"output": {
"result": {
"appliedProbability": 0.1,
"originalLoss": 4.605170185988091,
"flooredLoss": 2.3025850929940455,
"bias": -2.3025850929940455
}
}
}Limitations and assumptions
- The floor reduces or preserves this observed-event loss; no binary complement, probability-vector renormalization, or clipping policy is inferred.
- Outputs are nested inside result. 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
Keep clipping and normalization policies consistent across the complete probability model and evaluation set.
References
Similar or alternative tools
- Cross Entropy Reference Zero Support Flag
Validate both probabilities in [0, 1].
- Probability Quantization Surprisal Error
Return log2(original/quantized) for both probabilities in (0, 1].