Key Error Updating Data with selected Features

[dawbarn]

When running the cell:

2.6 Update dataset with selected features

I get 'KeyError: 'normal_e' at line 16. The error shows that I am trying to access a column named "normal_e", which is not present in the data.

I got it around the problem by implementing:

            if 'normal_e' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_e'] = []
            if 'normal_x' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_x'] = []

I applied this between under each of the if and elif statements, so it looked like:

def filter_features(data, gender=None):
    """
    gender: either None, 'm' or 'w'. Use m/w when trying out gender-aware model, this will add an extra node_feature
    the node feature will be 0 (m) and 1 (f). This is only relevent after loading both the M and W datasets (not implemented atm)
    """
    edge_feature_idxs = [idx for idx, x in enumerate(edge_f_box.children) if x.value]
    node_feature_idxs = [idx for idx, x in enumerate(node_f_box.children) if x.value]
    node_features = [x.description for x in node_f_box.children if x.value]

    # Check for empty edge features or node features.
    if not any(edge_feature_idxs) and not any(node_feature_idxs):
        print("\nCannot have zero edge features and zero node features.\n")
        print("\nDefaulting to the previous configuration.")    
    else:    
        if not gender:
            if 'normal_e' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_e'] = []
            if 'normal_x' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_x'] = []
            data[adj_matrix.value]['normal_e'] = [x[:, edge_feature_idxs] for x in data[adj_matrix.value]['normal_e']]
            data[adj_matrix.value]['normal_x'] = [x[:, node_feature_idxs] for x in data[adj_matrix.value]['normal_x']]
        elif gender == 'm':
            if 'normal_e' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_e'] = []
            if 'normal_x' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_x'] = []
            data[adj_matrix.value]['normal_e'] = [x[:, edge_feature_idxs] for x in data[adj_matrix.value]['normal_e']]
            data[adj_matrix.value]['normal_x'] = [np.append(
                x[:, node_feature_idxs], 
                np.zeros(shape=(x.shape[0], 1)), axis=1
            ) for x in data[adj_matrix.value]['normal_x']]
        elif gender == 'w':
            if 'normal_e' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_e'] = []
            if 'normal_x' not in data[adj_matrix.value]:
                data[adj_matrix.value]['normal_x'] = []
            data[adj_matrix.value]['normal_e'] = [x[:, edge_feature_idxs] for x in data[adj_matrix.value]['normal_e']]
            data[adj_matrix.value]['normal_x'] = [np.append(
                x[:, node_feature_idxs], 
                np.ones(shape=(x.shape[0], 1)), axis=1
            ) for x in data[adj_matrix.value]['normal_x']]
        else:
            raise NotImplementedError()
    
    return data

data = filter_features(og_data.copy())

However, I'm aware there may be another fix using .loc. Just a thought.