编写python代码设计一种基于模式挖掘的算法，要求挖掘出Cifar10数据集某一类中具有频繁性又有判别性的图片。

这是一个比较复杂的任务，需要使用多种技术来完成，以下是可能的实现步骤：

1. 加载Cifar10数据集，并选择一个类别进行挖掘。

import tensorflow as tf
from tensorflow.keras.datasets import cifar10

(train_images, train_labels), (test_images, test_labels) = cifar10.load_data()
class_name = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
selected_class = 'cat'
class_index = class_name.index(selected_class)
train_images = train_images[train_labels[:,0] == class_index]

2. 对于每个图片，提取其特征向量。

from tensorflow.keras.applications.resnet50 import ResNet50, preprocess_input
from tensorflow.keras.preprocessing.image import img_to_array, load_img

model = ResNet50(weights='imagenet', include_top=False)
features = []
for img in train_images:
    img = img_to_array(img)
    img = preprocess_input(img)
    feature = model.predict(img.reshape(1, 224, 224, 3))
    feature = feature.flatten()
    features.append(feature)

3. 使用Apriori算法挖掘频繁模式。

!pip install apyori
from apyori import apriori

min_support = 0.5
min_confidence = 0.7
rules = apriori(features, min_support=min_support, min_confidence=min_confidence, max_length=2)

4. 对挖掘出的频繁模式进行可视化。

import matplotlib.pyplot as plt
import numpy as np

support = []
confidence = []
for rule in rules:
    support.append(rule.support)
    confidence.append(rule.ordered_statistics[0].confidence)

plt.scatter(support, confidence, alpha=0.5)
plt.xlabel('support')
plt.ylabel('confidence')
plt.show()

5. 选择支持度和置信度较高的模式，对应的图片即为具有频繁性又有判别性的图片。

selected_rules = [rule for rule in rules if rule.support >= min_support and rule.ordered_statistics[0].confidence >= min_confidence]
selected_indices = set()
for rule in selected_rules:
    items = rule.items
    indices = [i for i in range(len(features)) if set(items).issubset(set(features[i]))]
    selected_indices.update(indices)

fig, axes = plt.subplots(1, len(selected_indices), figsize=(10, 10))
for i, idx in enumerate(selected_indices):
    img = train_images[idx]
    axes[i].imshow(img)
    axes[i].axis('off')
plt.show()

这里使用的是预训练的ResNet50模型提取特征向量，使用Apriori算法对特征向量进行挖掘，得到频繁模式后，选择支持度和置信度较高的模式对应的图片作为具有频繁性又有判别性的图片。可以根据实际需求修改参数和算法实现方式。