Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-2024-08-06· OpenAIAi2 WildChat-4.8M (AI2)

import numpy as np import matplotlib.pyplot as plt from sklearn.decomposition import PCA…

  1. User

    import numpy as np import matplotlib.pyplot as plt from sklearn.decomposition import PCA from sklearn.preprocessing import StandardScaler from sklearn.metrics.pairwise import cosine_similarity

    读取npy文件,并提取出相同标签对应的索引

    def process_labels(file_path): # 读取npy文件 labels = np.load(file_path)

    # 统计0, 1, 2对应的索引
    indices = {
        0: np.where(labels == 0)[0].tolist(),
        1: np.where(labels == 1)[0].tolist(),
        2: np.where(labels == 2)[0].tolist()
    }
    
    return indices
    

    根据相同标签的索引,取出对应的特征向量npy文件

    def process_features(file_path, indices, label, fe_num): index = indices[label] fea_list = [] # 遍历索引,取出对应的特征向量 for i in index: fea = np.load(f'{file_path}/{i}_fe{fe_num}.npy') fea_list.append(fea.squeeze(-2))

    return np.array(fea_list)
    

    def pca_similarity(sequence_vectors, n_components=None): num, seq, dim = sequence_vectors.shape

    # 1. 重塑序列向量为2D数组
    vectors = sequence_vectors.reshape(num, seq * dim)
    
    # 2. 标准化特征
    scaler = StandardScaler()
    scaled_vectors = scaler.fit_transform(vectors)
    
    # 3. PCA降维
    if n_components is None:
        n_components = min(num, seq * dim)
    
    pca = PCA(n_components=n_components)
    reduced_vectors = pca.fit_transform(scaled_vectors)
    
    # 4. 计算降维后向量的余弦相似度
    similarity_matrix = cosine_similarity(reduced_vectors)
    
    # 5. 计算平均相似度(排除自身与自身的相似度)
    total_similarity = (np.sum(similarity_matrix) - num) / (num * (num - 1))
    
    return total_similarity, pca.explained_variance_ratio_
    

    def plot_radar_chart(similarities, ax, title): categories = list(similarities.keys()) values = list(similarities.values())

    # 雷达图设置
    N = len(categories)
    angles = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist()
    values += values[:1]
    angles += angles[:1]
    
    ax.set_theta_offset(np.pi / 2)
    ax.set_theta_direction(-1)
    ax.set_rgrids([0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0], 
                 labels=['0','0.1', '0.2', '0.3', '0.4', '0.5', '0.6', '0.7', '0.8', '0.9', '1.0'])
    ax.set_rlabel_position(0)
    ax.plot(angles, values, linewidth=2, linestyle='solid')
    ax.fill(angles, values, alpha=0.25)
    
    ax.set_xticks(angles[:-1])
    ax.set_xticklabels(categories)
    ax.set_title(title)
    
    # 固定径向轴的范围
    ax.set_ylim(-0.1, 1)
    

    if name == "main": indices = process_labels(r'features_save_4\results.npy') fig, axes = plt.subplots(1, 3, subplot_kw=dict(polar=True), figsize=(24, 8)) # index = indices[2] # label = 2 # fea_1 = process_features('features_save_4', indices, label, 1) # total_similarity_1, _ = pca_similarity(fea_1) # fea_2 = process_features('features_save_4', indices, label, 2) # total_similarity_2, _ = pca_similarity(fea_2) # fea_3 = process_features('features_save_4', indices, label, 3) # total_similarity_3, _ = pca_similarity(fea_3) # fea_4 = process_features('features_save_4', indices, label, 4) # total_similarity_4, _ = pca_similarity(fea_4) # similarities = { # 'chest.Resp': total_similarity_1, # 'wrist.BVP': total_similarity_2, # 'chest.ECG': total_similarity_3, # 'wrist.EDA': total_similarity_4 # } # similarities = {k: v / np.sum(list(similarities.values())) for k, v in similarities.items()} # plot_radar_chart(similarities, axes[1], title=f'Label {label}')

    for i in range(3):
        index = indices[i]
        # 为空则跳过
        if len(index) == 0:
            label = i
            total_similarity_1 = 0
            total_similarity_2 = 0
            total_similarity_3 = 0
            total_similarity_4 = 0
            similarities = {
            'chest.Resp': total_similarity_1,
            'wrist.BVP': total_similarity_2,
            'chest.ECG': total_similarity_3,
            'wrist.EDA': total_similarity_4
            }
            plot_radar_chart(similarities, axes[i], title=f'Label {label}')
        else:    
            label = i
            fea_1 = process_features('features_save_4', indices, label, 1)
            total_similarity_1, _ = pca_similarity(fea_1)
            fea_2 = process_features('features_save_4', indices, label, 2)
            total_similarity_2, _ = pca_similarity(fea_2)
            fea_3 = process_features('features_save_4', indices, label, 3)
            total_similarity_3, _ = pca_similarity(fea_3)
            fea_4 = process_features('features_save_4', indices, label, 4)
            total_similarity_4, _ = pca_similarity(fea_4)
            similarities = {
            'chest.Resp': total_similarity_1,
            'wrist.BVP': total_similarity_2,
            'chest.ECG': total_similarity_3,
            'wrist.EDA': total_similarity_4
            }
            # 将similarities的各个值进行归一化
            similarities = {k: v / np.sum(list(similarities.values())) for k, v in similarities.items()}
        plot_radar_chart(similarities, axes[i], title=f'Label {label}')
    
    
    # 调整整体布局
    plt.tight_layout()  # 增加边距
    plt.savefig('radar_chart_4.png')  # 保存时确保边缘完整
    

    怎么调整轴标签和雷达图的距离

  2. Assistant

    要调整轴标签和雷达图之间的距离,可以通过 matplotlib 中的 tick_params 方法调整刻度标签的位置。特别是,我们可以设置 pad 参数来增加或减少标签和图形之间的距离。

    在 plot_radar_chart 函数中,可以添加以下行以调整标签与雷达图之间的距离:

    def plot_radar_chart(similarities, ax, title):
        categories = list(similarities.keys())
        values = list(similarities.values())
    
        # 雷达图设置
        N = len(categories)
        angles = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist()
        values += values[:1]
        angles += angles[:1]
    
        ax.set_theta_offset(np.pi / 2)
        ax.set_theta_direction(-1)
        ax.set_rgrids(
            [0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0],
            labels=['0', '0.1', '0.2', '0.3', '0.4', '0.5', '0.6', '0.7', '0.8', '0.9', '1.0']
        )
        ax.set_rlabel_position(0)
        ax.plot(angles, values, linewidth=2, linestyle='solid')
        ax.fill(angles, values, alpha=0.25)
    
        ax.set_xticks(angles[:-1])
        ax.set_xticklabels(categories)
        ax.set_title(title)
    
        # Adjusting the label distance
        ax.xaxis.set_tick_params(pad=10)  # Increase the distance between the labels and the plot
        ax.yaxis.set_tick_params(pad=10)  # Adjust this for radial labels, if needed
    
        # 固定径向轴的范围
        ax.set_ylim(-0.1, 1)
    

    在这段代码中,ax.xaxis.set_tick_params(pad=10) 是用于调整轴标签(特定类别标签)与雷达图之间的距离的行。你可以根据你的需求调整 pad 的值来控制距离。请记得,这个距离的数值以点为单位,并且不同系统上的视觉效果可能会略有不同,具体数值可以根据实际需要进行尝试和调整。