Utilizando KERAS - Red Neuronal MNIST Para principiantes - TensorFlow 2 quickstart for beginners con Keras

 Red Neuronal MNIST Para principiantes - TensorFlow 2 quickstart for beginners

Red Neuronal MNIST Para principiantes

El mejor punto de partida es la API secuencial, que es fácil de usar. Puedes crear modelos si conectas bloques de compilación. Ejecuta el ejemplo de "Hola, mundo" que se encuentra a continuación y, luego, consulta los instructivos para obtener más información.

Para aprender sobre el AA, visita nuestra página educativa. Comienza con capacitaciones seleccionadas para mejorar tus habilidades en áreas fundamentales del AA.

https://www.tensorflow.org/overview

import tensorflow as tf
mnist = tf.keras.datasets.mnist

(x_train, y_train),(x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(28, 28)),
  tf.keras.layers.Dense(128, activation='relu'),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10, activation='softmax')
])

model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

model.fit(x_train, y_train, epochs=5)
model.evaluate(x_test, y_test)

TensorFlow 2 quickstart for beginners

https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/quickstart/beginner.ipynb#scrollTo=3wF5wszaj97Y

This short introduction uses Keras to:

1. Build a neural network that classifies images.
2. Train this neural network.
3. And, finally, evaluate the accuracy of the model.

---

This is a Google Colaboratory notebook file. Python programs are run directly in the browser—a great way to learn and use TensorFlow. To follow this tutorial, run the notebook in Google Colab by clicking the button at the top of this page.

1. In Colab, connect to a Python runtime: At the top-right of the menu bar, select CONNECT.
2. Run all the notebook code cells: Select Runtime > Run all.

---

Download and install TensorFlow 2. Import TensorFlow into your program:

Note: Upgrade pip to install the TensorFlow 2 package. See the install guide for details.

---

[2]

import tensorflow as tf

---

Load and prepare the MNIST dataset. Convert the samples from integers to floating-point numbers:

---

[3]

mnist = tf.keras.datasets.mnist

 

(x_train, y_train), (x_test, y_test) = mnist.load_data()

x_train, x_test = x_train / 255.0, x_test / 255.0

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11493376/11490434 [==============================] - 0s 0us/step

---

Build the tf.keras.Sequential model by stacking layers. Choose an optimizer and loss function for training:

---

[4]

model = tf.keras.models.Sequential([

  tf.keras.layers.Flatten(input_shape=(28, 28)),

  tf.keras.layers.Dense(128, activation='relu'),

  tf.keras.layers.Dropout(0.2),

  tf.keras.layers.Dense(10)

])

---

For each example the model returns a vector of "logits" or "log-odds" scores, one for each class.

---

[5]

predictions = model(x_train[:1]).numpy()

predictions

array([[-0.10128427, -0.14768334,  0.61712044,  0.61311674, -0.25687024,
        -0.24292384,  0.23817754, -0.5163672 , -0.28512543, -0.35918558]],
      dtype=float32)

---

The tf.nn.softmax function converts these logits to "probabilities" for each class:

---

[6]

tf.nn.softmax(predictions).numpy()

array([[0.08739848, 0.08343592, 0.17926814, 0.17855184, 0.07480554,
        0.07585612, 0.12272422, 0.05770795, 0.07272148, 0.06753031]],
      dtype=float32)

---

Note: It is possible to bake this tf.nn.softmax in as the activation function for the last layer of the network. While this can make the model output more directly interpretable, this approach is discouraged as it's impossible to provide an exact and numerically stable loss calculation for all models when using a softmax output.

---

The losses.SparseCategoricalCrossentropy loss takes a vector of logits and a True index and returns a scalar loss for each example.

---

[7]

loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)

---

This loss is equal to the negative log probability of the true class: It is zero if the model is sure of the correct class.

This untrained model gives probabilities close to random (1/10 for each class), so the initial loss should be close to -tf.log(1/10) ~= 2.3.

---

[8]

loss_fn(y_train[:1], predictions).numpy()

2.578917

---

[9]

model.compile(optimizer='adam',

              loss=loss_fn,

              metrics=['accuracy'])

---

The Model.fit method adjusts the model parameters to minimize the loss:

---

[10]

model.fit(x_train, y_train, epochs=5)

Epoch 1/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.2970 - accuracy: 0.9146
Epoch 2/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.1473 - accuracy: 0.9563
Epoch 3/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.1096 - accuracy: 0.9671
Epoch 4/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.0897 - accuracy: 0.9719
Epoch 5/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.0752 - accuracy: 0.9760

<tensorflow.python.keras.callbacks.History at 0x7fd755cb47f0>

---

The Model.evaluate method checks the models performance, usually on a "Validation-set" or "Test-set".

---

[11]

model.evaluate(x_test,  y_test, verbose=2)

313/313 - 0s - loss: 0.0770 - accuracy: 0.9775

[0.07695907354354858, 0.9775000214576721]

---

The image classifier is now trained to ~98% accuracy on this dataset. To learn more, read the TensorFlow tutorials.

---

If you want your model to return a probability, you can wrap the trained model, and attach the softmax to it:

---

[12]

probability_model = tf.keras.Sequential([

  model,

  tf.keras.layers.Softmax()

])

---

[13]

probability_model(x_test[:5])

<tf.Tensor: shape=(5, 10), dtype=float32, numpy=
array([[2.1094127e-08, 1.3406939e-08, 2.6930111e-06, 1.0430716e-04,
        1.1646294e-10, 7.5275358e-07, 4.9344961e-14, 9.9987352e-01,
        4.5862681e-08, 1.8621193e-05],
       [6.6366977e-07, 6.6512353e-06, 9.9998164e-01, 1.4112009e-06,
        8.1197561e-15, 4.9681449e-07, 1.4861804e-06, 4.9511540e-13,
        7.6379356e-06, 4.1590975e-14],
       [1.5588105e-06, 9.9926704e-01, 8.1892227e-05, 2.6095991e-05,
        1.2973814e-05, 5.7134803e-06, 2.4758489e-05, 4.2972015e-04,
        1.4964781e-04, 6.4341845e-07],
       [9.9977213e-01, 1.5741728e-09, 7.8205092e-05, 1.2229310e-07,
        9.9353588e-07, 2.4155740e-06, 1.3811085e-04, 1.5705807e-06,
        2.3177182e-07, 6.2728159e-06],
       [1.7277871e-05, 1.0273905e-09, 4.3157845e-05, 3.0481058e-08,
        9.9705887e-01, 6.6911116e-06, 8.1449107e-06, 4.3538297e-05,
        2.4825749e-06, 2.8197672e-03]], dtype=float32)>