| # | Likes | Tech tags | Title | Creator | Created date |
|---|---|---|---|---|---|
| 1 | 0 |
TensorFlow
Keras
|
2022-10-11 17:56
|
Creates a simple 3-layer feedforward Neural Network with 500 and 150 hidden units in the hidden layers. Uses sigmoid activation functions.
It achieves 96.1% accuracy on the MNIST test set.
It requires about 30s to train on the CPU of the personal computer it was developed on.
from tensorflow import keras
# Seed the Pseudorandom number generators to make the code deterministic
keras.utils.set_random_seed(0)
# Load the dataset
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data(path="mnist.npz")
y_train = keras.utils.to_categorical(y_train, 10)
y_test = keras.utils.to_categorical(y_test, 10)
# Construct the neural network
inputs = keras.Input(shape=x_train.shape[1:])
hidden = keras.layers.Flatten()(inputs)
hidden = keras.layers.Dense(500, activation='sigmoid')(hidden)
hidden = keras.layers.Dense(150, activation='sigmoid')(hidden)
outputs = keras.layers.Dense(10, activation='softmax')(hidden)
model = keras.Model(inputs, outputs, name="simple_feedforward_sample")
model.compile(
optimizer=keras.optimizers.RMSprop(learning_rate=1e-3),
loss=keras.losses.CategoricalCrossentropy(),
metrics=["acc"]
)
# Train the model on the training dataset
model.fit(x_train, y_train, batch_size=64, epochs=7, validation_split=0.2)
# Evaluate the accuracy of the model
model.evaluate(x_test, y_test)
| classes | |
tensorflow.keras.Input |
tensorflow.org |
tensorflow.keras.Model |
tensorflow.org |
tensorflow.keras.layers.Dense |
tensorflow.org |
tensorflow.keras.layers.Flatten |
tensorflow.org |
tensorflow.keras.losses.CategoricalCrossentropy |
tensorflow.org |
tensorflow.keras.optimizers.RMSprop |
tensorflow.org |
| functions | |
tensorflow.keras.Model.compile |
tensorflow.org |
tensorflow.keras.Model.evaluate |
tensorflow.org |
tensorflow.keras.Model.fit |
tensorflow.org |
tensorflow.keras.datasets.mnist.load_data |
tensorflow.org |
tensorflow.keras.utils.set_random_seed |
tensorflow.org |
tensorflow.keras.utils.to_categorical |
tensorflow.org |
Create an algorithm that classifies the images in the MNIST dataset test set into the digits they correspond to.
You are allowed to use a training dataset to train your algorithm, but the training dataset should be separate from the test dataset.
Your algorithm should strive to maximize accuracy, and you should specify what accuracy you achieve. Your algorithm should also be generalizable, that is work well on the type of digit images provided in MNIST, even if they don't happen to be in the test set.