吴恩达 Andrew Ng. Machine Learning C2_W3_Assignment Practice Lab

Exercise 1

# UNQ_C1

# GRADED CELL: eval_mse

def eval_mse(y, yhat):

    """

    Calculate the mean squared error on a data set.

    Args:

      y    : (ndarray  Shape (m,) or (m,1))  target value of each example

      yhat : (ndarray  Shape (m,) or (m,1))  predicted value of each example

    Returns:

      err: (scalar)            

    """

    m = len(y)

    err = 0.0

    for i in range(m):

    ### START CODE HERE ###

        err+=(y[i]-yhat[i])**2

    err=err/(2*m)

    ### END CODE HERE ###

   

    return(err)

 

Exercise 2

# UNQ_C2

# GRADED CELL: eval_cat_err

def eval_cat_err(y, yhat):

    """

    Calculate the categorization error

    Args:

      y    : (ndarray  Shape (m,) or (m,1))  target value of each example

      yhat : (ndarray  Shape (m,) or (m,1))  predicted value of each example

    Returns:|

      cerr: (scalar)            

    """

    m = len(y)

    incorrect = 0

    for i in range(m):

    ### START CODE HERE ###

        if y[i]!=yhat[i]:

            incorrect+=1

    cerr=incorrect/m

    ### END CODE HERE ###

   

    return(cerr)

Exercise 3

# UNQ_C3

# GRADED CELL: model

import logging

logging.getLogger("tensorflow").setLevel(logging.ERROR)

tf.random.set_seed(1234)

model = Sequential(

    [

        ### START CODE HERE ###

      Dense(units=120,activation='relu'),

      Dense(units=40,activation='relu'),

      Dense(units=6,activation='linear'),

        ### END CODE HERE ###

    ], name="Complex"

)

model.compile(

    ### START CODE HERE ###

    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),

    optimizer=tf.keras.optimizers.Adam(learning_rate=0.01)

    ### END CODE HERE ###

)

Exercise 4

# UNQ_C4

# GRADED CELL: model_s

tf.random.set_seed(1234)

model_s = Sequential(

    [

        ### START CODE HERE ###

      Dense(units=6,activation='relu'),

      Dense(units=6,activation='linear'),

        ### END CODE HERE ###

    ], name = "Simple"

)

model_s.compile(

    ### START CODE HERE ###

    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),

    optimizer=tf.keras.optimizers.Adam(learning_rate=0.01)

    ### START CODE HERE ###

)

Exercise 5

# UNQ_C5

# GRADED CELL: model_r

tf.random.set_seed(1234)

model_r = Sequential(

    [

        ### START CODE HERE ###

        Dense(units=120,activation='relu',kernel_regularizer=tf.keras.regularizers.l2(0.1),name="L1"),

        Dense(units=40,activation='relu',kernel_regularizer=tf.keras.regularizers.l2(0.1),name="L2"),

        Dense(units=6,activation='linear',name="L3"),

        ### START CODE HERE ###

    ], name= None

)

model_r.compile(

    ### START CODE HERE ###

    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),

    optimizer=tf.keras.optimizers.Adam(learning_rate=0.01)

    ### START CODE HERE ###

)