需要帮助来理解这一行代码（字典，键，熊猫，numpy）

Question

I'm attempting a Google Crash Course to learn TensorFlow and Machine Learning. I am having trouble comprehending one of the lines from their coding examples .

def my_input_fn(features, targets, batch_size=1, shuffle=True, num_epochs=None):
"""Trains a linear regression model of one feature.

Args:
  features: pandas DataFrame of features
  targets: pandas DataFrame of targets
  batch_size: Size of batches to be passed to the model
  shuffle: True or False. Whether to shuffle the data.
  num_epochs: Number of epochs for which data should be repeated. None = repeat indefinitely
Returns:
  Tuple of (features, labels) for next data batch
"""

# Convert pandas data into a dict of np arrays.
features = {key:np.array(value) for key,value in dict(features).items()}

---

I need help understanding that last line of code.

features = {key:np.array(value) for key,value in dict(features).items()}

---

I've researched dictionaries in an attempt to understand it myself, but it's still a bit much for me to grasp. I've attempted to write the same line of code in a way I can understand:

np_dict_array = dict(features).items()

for key,value in np_dict_array:
    features += np_dict_array[key]

I do not think I am rewriting the code correctly. To get specific, I need help understanding what this does in that line of code:

key:np.array(value)

---

If anybody could explain what that line of code is doing, or (bonus points) rewrite it in a novice-friendly way, I would greatly appreciate it!

Answer 1

features = {key:np.array(value) for key,value in dict(features).items()}

It is a dictionary comprehension . It converts all values in dict(features) to a Numpy array.

key:np.array(value)

This is how you assign key value pairs to a dictionary.

Answer 2

Alternate syntax:

features = {}
for key, value in dict(features).items():
    features[key] = np.array(value)

Comprehensions are popular as they reduce this sort of common pattern down to a single line. However, it is sometimes tempting to try to do too much in a comprehension as complexity grows.

Answer 3

This is a 'dictionary comprehension' - modelled on a list comprehension, but making a new dictionary instead.

features = {key:np.array(value) for key,value in dict(features).items()}

take things from the inside out:

dict(features)    # make a dictionary from the `features` argument
    .items()      # make a list of (key,value) tuples
for key,value     # iterate on those tuples
np.array(value)   # make a numpy array from the value
key:...           # make a new entry in the new dictionary

In sum, it makes a dictionary from features , making sure that the value for each item is a numpy array.

fdict = dict(features)
adict = dict()   # empty dictionary
for key,value in fdict.items():
    adict[key] = np.array(value)