eXNN.InnerNeuralTopology.api 

 1import matplotlib
 2import torch
 3from typing import Dict, List
 4from eXNN.InnerNeuralTopology.homologies import InnerNetspaceHomologies, _ComputeBarcode
 5
 6
 7def ComputeBarcode(data: torch.Tensor,
 8                   hom_type: str,
 9                   coefs_type: str) -> matplotlib.figure.Figure:
10    """This function plots persistent homologies for a cloud of points as barcodes.
11
12    Args:
13        data (torch.Tensor): input data of shape NxC1x...xCk,
14            where N is the number of data points,
15            C1,...,Ck are dimensions of each data point
16        hom_type (str): homotopy type
17        coefs_type (str): coefficients type
18
19    Returns:
20        matplotlib.figure.Figure: barcode plot
21    """
22    return _ComputeBarcode(data, hom_type, coefs_type)
23
24
25def NetworkHomologies(model: torch.nn.Module,
26                      data: torch.Tensor,
27                      layers: List[str],
28                      hom_type: str,
29                      coefs_type: str) -> Dict[str, matplotlib.figure.Figure]:
30    """
31    The function plots persistent homologies for latent representations
32        on different levels of the neural network as barcodes.
33
34    Args:
35        model (torch.nn.Module): neural network
36        data (torch.Tensor): input data of shape NxC1x...xCk,
37            where N is the number of data points,
38            C1,...,Ck are dimensions of each data point
39        layers (List[str]): list of layers for visualization. Defaults to None.
40            If None, visualization for all layers is performed
41        hom_type (str): homotopy type
42        coefs_type (str): coefficients type
43
44    Returns:
45        Dict[str, matplotlib.figure.Figure]: dictionary with a barcode plot for each layer
46    """
47    res = {}
48    for layer in layers:
49        res[layer] = InnerNetspaceHomologies(model, data, layer, hom_type, coefs_type)
50    return res
def ComputeBarcode( data: torch.Tensor, hom_type: str, coefs_type: str) -> matplotlib.figure.Figure: 
 8def ComputeBarcode(data: torch.Tensor,
 9                   hom_type: str,
10                   coefs_type: str) -> matplotlib.figure.Figure:
11    """This function plots persistent homologies for a cloud of points as barcodes.
12
13    Args:
14        data (torch.Tensor): input data of shape NxC1x...xCk,
15            where N is the number of data points,
16            C1,...,Ck are dimensions of each data point
17        hom_type (str): homotopy type
18        coefs_type (str): coefficients type
19
20    Returns:
21        matplotlib.figure.Figure: barcode plot
22    """
23    return _ComputeBarcode(data, hom_type, coefs_type)

This function plots persistent homologies for a cloud of points as barcodes.

Args: data (torch.Tensor): input data of shape NxC1x...xCk, where N is the number of data points, C1,...,Ck are dimensions of each data point hom_type (str): homotopy type coefs_type (str): coefficients type

Returns: matplotlib.figure.Figure: barcode plot

def NetworkHomologies( model: torch.nn.modules.module.Module, data: torch.Tensor, layers: List[str], hom_type: str, coefs_type: str) -> Dict[str, matplotlib.figure.Figure]: 
26def NetworkHomologies(model: torch.nn.Module,
27                      data: torch.Tensor,
28                      layers: List[str],
29                      hom_type: str,
30                      coefs_type: str) -> Dict[str, matplotlib.figure.Figure]:
31    """
32    The function plots persistent homologies for latent representations
33        on different levels of the neural network as barcodes.
34
35    Args:
36        model (torch.nn.Module): neural network
37        data (torch.Tensor): input data of shape NxC1x...xCk,
38            where N is the number of data points,
39            C1,...,Ck are dimensions of each data point
40        layers (List[str]): list of layers for visualization. Defaults to None.
41            If None, visualization for all layers is performed
42        hom_type (str): homotopy type
43        coefs_type (str): coefficients type
44
45    Returns:
46        Dict[str, matplotlib.figure.Figure]: dictionary with a barcode plot for each layer
47    """
48    res = {}
49    for layer in layers:
50        res[layer] = InnerNetspaceHomologies(model, data, layer, hom_type, coefs_type)
51    return res

The function plots persistent homologies for latent representations on different levels of the neural network as barcodes.

Args: model (torch.nn.Module): neural network data (torch.Tensor): input data of shape NxC1x...xCk, where N is the number of data points, C1,...,Ck are dimensions of each data point layers (List[str]): list of layers for visualization. Defaults to None. If None, visualization for all layers is performed hom_type (str): homotopy type coefs_type (str): coefficients type

Returns: Dict[str, matplotlib.figure.Figure]: dictionary with a barcode plot for each layer