如何针对 TF.Dataset 的 class label 的 plot 直方图

Question

I'm loading data from the disk with TF.CsvDataset . And plotting the data as

#This is the transformation function applied on loaded data before displaying histogram.
def preprocess(*fields):
    print(len(fields))
    features=tf.stack(fields[:-1])
    labels=tf.stack([int(x) for x in fields[-1:]])
    return features,labels  # x, y

for features,label in train_ds.take(1000):
#  print(features[0])
plt.hist(features.numpy().flatten(), bins = 101)

And I'm getting this histogram

But I want to plot distribution of 712 features' values against binary class labels. That is, what is the value of feature 1,2 or 3 when class label is 0.

How to do that with pyplot?

I have read following threads but, nothing helped.

Plotting histograms against classes in pandas / matplotlib

Histogram color by class

How to draw an histogram with multiple categories in python

Answer 1

You can use np.fromiter and get all the labels. Then you simply pass the list of labels to plt.hist :

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

train, test = tf.keras.datasets.mnist.load_data()

ds = tf.data.Dataset.from_tensor_slices(train)

vals = np.fromiter(ds.map(lambda x, y: y), float)

plt.hist(vals)
plt.xticks(range(10))
plt.title('Label Frequency')
plt.show()