Inputs for a numerical gradient check
Reference inputs · synthetic data
- Records
- 3
- Data columns
- 3
- Missing cells
- 0
- Duplicate records
- 0
- Feature duplicates
- 0
3 of 3 records · One record = one evaluation point. Statistics and downloads use all records.
About this dataset
These are the center and two nearby inputs used to evaluate z = (x + y)x. They make the finite-difference comparison inspectable before calculating the gradient.
- Three deterministic probes, not three independent training samples.
- At very small h, float64 may store a perturbed x as the same value as the center. Feature duplicates can reveal that loss of resolution.
- Changing the backward-path setting leaves these inputs and the forward function unchanged.
Counts exclude the row identifier. Duplicate counts are extra occurrences after the first. Feature duplicates ignore labels and experimental conditions. — means a statistic is undefined or not applicable.
- Row
- Stable identifier within this synthetic fixture.
- Evaluation
- Center, left, or right probe; these are dependent evaluation points.
- Input x
- The varied input, stored as a float64 number.
- Input y
- Held fixed during the gradient check.