1 MonotoneBinValues

rri_metrics.coral.MonotoneBinValues(
    num_classes,
    init_values,
    *,
    min_delta=1e-06,
)

Learnable, monotone bin representatives u_k.

We parameterize u_k via a base value and positive deltas:

u_0 in R
u_k = u_0 + sum_{j=1..k} softplus(delta_j)

This guarantees u_0 <= u_1 <= ... <= u_{K-1} while keeping gradients stable.

1.1 Methods

Name	Description
values	Return monotone bin representatives `u_k`.
reset_from_values	Re-initialize from target values (keeps monotonicity).

1.1.1 values

rri_metrics.coral.MonotoneBinValues.values()

Return monotone bin representatives u_k.

1.1.2 reset_from_values

rri_metrics.coral.MonotoneBinValues.reset_from_values(
    values,
    *,
    min_delta=1e-06,
)

Re-initialize from target values (keeps monotonicity).

# MonotoneBinValues { #aria_nbv.rri_metrics.coral.MonotoneBinValues }

```python
rri_metrics.coral.MonotoneBinValues(
    num_classes,
    init_values,
    *,
    min_delta=1e-06,
)
```

Learnable, monotone bin representatives ``u_k``.

We parameterize ``u_k`` via a base value and positive deltas:

    u_0 in R
    u_k = u_0 + sum_{j=1..k} softplus(delta_j)

This guarantees ``u_0 <= u_1 <= ... <= u_{K-1}`` while keeping gradients stable.

## Methods

| Name | Description |
| --- | --- |
| [values](#aria_nbv.rri_metrics.coral.MonotoneBinValues.values) | Return monotone bin representatives ``u_k``. |
| [reset_from_values](#aria_nbv.rri_metrics.coral.MonotoneBinValues.reset_from_values) | Re-initialize from target values (keeps monotonicity). |

### values { #aria_nbv.rri_metrics.coral.MonotoneBinValues.values }

```python
rri_metrics.coral.MonotoneBinValues.values()
```

Return monotone bin representatives ``u_k``.

### reset_from_values { #aria_nbv.rri_metrics.coral.MonotoneBinValues.reset_from_values }

```python
rri_metrics.coral.MonotoneBinValues.reset_from_values(
    values,
    *,
    min_delta=1e-06,
)
```

Re-initialize from target values (keeps monotonicity).