Experimenting With Models
Build. Fit. Evaluate. ReBuild. ReFit. ReEvaluate. ReBuild. ReFit...
- use more data
- change number of layers
- change the number of "units" in each layer
- train for longer
Imports
In [1]:
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
import pandas as pdMock Some Data
In [2]:
X = np.arange(-100, 100, 4)
y = np.arange(-90,110,4)Split Data Into Training + Testing
In [3]:
# Split data into train and test sets
X_train = X[:40] # first 40 examples (80% of data)
y_train = y[:40]
X_test = X[40:] # last 10 examples (20% of data)
y_test = y[40:]Visualize
In [4]:
plt.figure(figsize=(10, 7))
# Plot training data in blue
plt.scatter(X_train, y_train, c='b', label='Training data')
# Plot test data in green
plt.scatter(X_test, y_test, c='g', label='Testing data')
# Show the legend
plt.legend();
Begin Experimenting
Model 1
In [5]:
# for repeatability
tf.random.set_seed(42)
m1Epochs = 100
moreEpochs = 500
lessEpochs = 25
fittedXTrained = tf.expand_dims(X_train, axis=-1)
# BUILD
m1 = tf.keras.Sequential()
l1 = tf.keras.layers.Dense(1)
m1.add(l1)
# COMPILE
m1.compile(loss=tf.keras.losses.mae,
optimizer=tf.keras.optimizers.SGD(),
metrics=['mae'])
# FIT
m1.fit(fittedXTrained,y_train, epochs=m1Epochs)Out [5]:
Predict some outputs
In [6]:
m1Predictions = m1.predict(X_test)In [7]:
def plot_predictions(train_data=X_train,
train_labels=y_train,
test_data=X_test,
test_labels=y_test,
predictions=y_test):
"""
Plots training data, test data and compares predictions.
"""
plt.figure(figsize=(10, 7))
# Plot training data in blue
plt.scatter(train_data, train_labels, c="b", label="Training data")
# Plot test data in green
plt.scatter(test_data, test_labels, c="g", label="Testing data")
# Plot the predictions in red (predictions were made on the test data)
plt.scatter(test_data, predictions, c="r", label="Predictions")
# Show the legend
plt.legend();
def mae(y_test, y_pred):
"""
Calculuates mean absolute error between y_test and y_preds.
"""
return tf.metrics.mean_absolute_error(y_test,
y_pred)
def mse(y_test, y_pred):
"""
Calculates mean squared error between y_test and y_preds.
"""
return tf.metrics.mean_squared_error(y_test,
y_pred)Visualize Test, Train, And Predictions
In [8]:
plot_predictions(predictions=m1Predictions)
Calculate some model Stats
In [9]:
# Calculate model_1 metrics
mae_1 = mae(y_test, m1Predictions.squeeze()).numpy()
mse_1 = mse(y_test, m1Predictions.squeeze()).numpy()
mae_1, mse_1Out [9]:
Model 2: More Layers
Build
In [10]:
# BUILD
m2 = tf.keras.Sequential()
m2.add(l1)
m2.add(l1)
# COMPILE
m2.compile(loss=tf.keras.losses.mae,
optimizer=tf.keras.optimizers.SGD(),
metrics=['mae'])
# FIT
m2.fit(fittedXTrained, y_train, epochs=m1Epochs, verbose=0) # set verbose to 0 for less outputOut [10]:
Predict
In [11]:
predictionsM2 = m2.predict(X_test)
plot_predictions(predictions=predictionsM2)
Get Model Stats
In [12]:
mae_2 = mae(y_test, predictionsM2.squeeze()).numpy()
mse_2 = mse(y_test, predictionsM2.squeeze()).numpy()
mae_2, mse_2Out [12]:
Model 3: More Epochs
Build
In [13]:
# BUILD
m3 = tf.keras.Sequential()
m3.add(l1)
m3.add(l1)
# COMPILE
m3.compile(loss=tf.keras.losses.mae,
optimizer=tf.keras.optimizers.SGD(),
metrics=['mae'])
# FIT
m3.fit(fittedXTrained, y_train, epochs=moreEpochs, verbose=0) # set verbose to 0 for less outputOut [13]:
Predict
In [14]:
predictionsM3 = m3.predict(X_test)
plot_predictions(predictions=predictionsM3)
Get Model Stats
In [15]:
mae_3 = mae(y_test, predictionsM3.squeeze()).numpy()
mse_3 = mse(y_test, predictionsM3.squeeze()).numpy()
mae_3, mse_3Out [15]:
Model 4: less epochs
In [18]:
# BUILD
m4 = tf.keras.Sequential()
m4.add(l1)
m4.add(l1)
# COMPILE
m4.compile(loss=tf.keras.losses.mae,
optimizer=tf.keras.optimizers.SGD(),
metrics=['mae'])
# FIT
m4.fit(fittedXTrained, y_train, epochs=lessEpochs, verbose=0) # set verbose to 0 for less output
predictionsM4 = m4.predict(X_test)
plot_predictions(predictions=predictionsM4)
mae_4 = mae(y_test, predictionsM4.squeeze()).numpy()
mse_4 = mse(y_test, predictionsM4.squeeze()).numpy()
mae_4, mse_4Out [18]:
In [19]:
resultsArray = [["model_1", mae_1, mse_1],
["model_2", mae_2, mse_2],
["model_3", mae_3, mae_3],
["model_4", mae_4, mae_4]
]
all_results = pd.DataFrame(resultsArray, columns=["model", "mae", "mse"])
all_resultsOut [19]: