IsoScore

https://pypi.org/project/IsoScore/
Admin
IsoScore is a tool that accurately measures isotropy in embedding space, with an updated version...
Tokens:847
Snippets:13
Trust Score:10
Update:5 months ago
Show doc for...
Tokens:
Raw
### IsoScore Functions

Source: https://pypi.org/project/IsoScore

This section details the functions available for use after installing IsoScore via pip.

```APIDOC
## Python Usage

Import the necessary components from the IsoScore library:
```python
import torch
from IsoScore.IsoScore import *
```

### Available Functions:

*   **`IsoScore(points)`**
    *   Description: Measures the isotropy of a given point cloud.
    *   Parameters:
        *   `points` (torch.Tensor): The input point cloud data.

*   **`IsoScore_star(points, C0, zeta)`**
    *   Description: Calculates IsoScore* for point clouds where dimensionality exceeds the number of samples. This version is differentiable and can be used as a regularizer.
    *   Parameters:
        *   `points` (torch.Tensor): The input point cloud data.
        *   `C0` (torch.Tensor): The shrinkage covariance matrix.
        *   `zeta` (float): The shrinkage parameter.

*   **`IsoScore_cov(cov)`**
    *   Description: Calculates IsoScore directly from a covariance matrix, useful when sample generation is not needed.
    *   Parameters:
        *   `cov` (torch.Tensor): The covariance matrix.

*   **`istar()`**
    *   Description: A `torch.nn.Module` implementation for the I-STAR loss.
```

--------------------------------

### I-STAR Loss Module

Source: https://pypi.org/project/IsoScore

Utilize the `istar()` function to get a `torch.nn.Module` for implementing the I-STAR loss. This module can be integrated into PyTorch training pipelines.

```python
istar()
```

--------------------------------

### Import IsoScore Functions

Source: https://pypi.org/project/IsoScore

Import necessary functions from the IsoScore library. This is typically the first step before using any IsoScore functionality.

```python
import torch
from IsoScore.IsoScore import *

```

--------------------------------

### Calculate IsoScore*

Source: https://pypi.org/project/IsoScore

Calculate the updated IsoScore* for point clouds, especially useful when the dimensionality exceeds the number of samples. Requires the shrinkage covariance matrix (C0) and shrinkage parameter (zeta).

```python
IsoScore_star(points, C0, zeta)
```

--------------------------------

### Calculate Original IsoScore

Source: https://pypi.org/project/IsoScore

Use the original IsoScore function to measure the isotropy of a given point cloud. Ensure your points are represented as a PyTorch tensor.

```python
IsoScore(points)
```

--------------------------------

### Calculate IsoScore from Covariance Matrix

Source: https://pypi.org/project/IsoScore

Directly calculate IsoScore from a given covariance matrix without needing to generate samples. This is useful for analyzing the isotropy of a known distribution.

```python
IsoScore_cov(cov)
```
Tokens:
Raw