Download as PDF, PPTX
![MNIST training data
>>> import matplotlib.pyplot as plt
>>> plt.imshow(train_images[0])
>>> plt.gray()
>>> plt.show()
>>> train_labels
array([5, 0, 4, ..., 5, 6, 8], dtype=uint8)](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-19-2048.jpg)
![MNIST testing data
>>> test_images.shape
(10000, 28, 28)
>>> test_labels
array([7, 2, 1, ..., 4, 5, 6], dtype=uint8)
Testing input: 10,000 images (28 pixels * 28 pixels)
Testing labels: 10,000
Training
60,000 images
60,000 labels
Testing
10,000 images
10,000 labels](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-20-2048.jpg)
![Data Reshaping (Input X)
# Train images
>>> train_images = train_images.reshape((60000, 28 * 28))
>>> train_images = train_images.astype('float32') / 255
# Test images
>>> test_images = test_images.reshape((10000, 28 * 28))
>>> test_images = test_images.astype('float32') / 255
1. 28 * 28 image (int values 0 -
255)
2. 28 * 28 = 784-element vector:
[row 0, row 1, …, row 27]
3. 784-element vectors with float
values in range 0 to 1](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-21-2048.jpg)
![Data Reshaping (Labels Y)
>>> from keras.utils import to_categorical
# Train labels
>>> train_labels = to_categorical(train_labels)
# Test labels
>>> test_labels = to_categorical(test_labels)
1. train_labels = [5 0 4 ... 5 6 8]
2. train_labels =
[[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]
[1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]
…
[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]
[0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
[0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]]](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-22-2048.jpg)
![Network compilation
>>> network.compile(optimizer='rmsprop',
... loss='categorical_crossentropy',
... metrics=['accuracy'])](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-26-2048.jpg)
![Let’s train!
>>> network.fit(train_images, train_labels, epochs=5, batch_size=128)
Epoch 1/5
60000/60000 [==============================] - 2s 33us/step - loss: 0.2179 - acc: 0.9329
Epoch 2/5
60000/60000 [==============================] - 2s 31us/step - loss: 0.0805 - acc: 0.9750
Epoch 3/5
60000/60000 [==============================] - 2s 32us/step - loss: 0.0531 - acc: 0.9841
Epoch 4/5
60000/60000 [==============================] - 2s 31us/step - loss: 0.0385 - acc: 0.9883
Epoch 5/5
60000/60000 [==============================] - 2s 31us/step - loss: 0.0290 - acc: 0.9908
Train
Accuracy: 99.08%](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-28-2048.jpg)
![Test Accuracy
>>> test_loss, test_acc = network.evaluate(test_images, test_labels)
10000/10000 [==============================] - 0s 47us/step
>>> print('Test accuracy: %s' % test_acc)
Test accuracy: 0.9799
Test Accuracy: 97.99%](https://image.slidesharecdn.com/introductiontodeeplearningusingpython-180516210151/75/Introduction-to-deep-learning-using-python-29-2048.jpg)
Uploaded byLino Coria
Introduction to deep learning using python
This document introduces deep learning using Python, focusing on the practical application of deep neural networks with the Keras library. It provides a tutorial on how to implement a neural network to classify the MNIST dataset of handwritten digits, detailing the process from data preprocessing to network training and evaluation. Additionally, it highlights key resources and platforms for learning deep learning concepts and techniques.
More Related Content
Similar to Introduction to deep learning using python
![[DSC Europe 25] Max Talanov - Non digital NNs.pptx](https://cdn.slidesharecdn.com/ss_thumbnails/wif8tr3gtua74qvtopke-non-digital-nns-251205090438-26b0eea6-thumbnail.jpg?width=640&height=640&fit=bounds)
![[DSC Europe 25] Boris Perkovic - Lost in performance.pptx](https://cdn.slidesharecdn.com/ss_thumbnails/uq5hrp7vsuahqkxzifux-1-251204082258-fd2ee09d-thumbnail.jpg?width=640&height=640&fit=bounds)
![[DSC Europe 25] Jim Sterne - Adopting Generative AI Capabilities Into the Ent...](https://cdn.slidesharecdn.com/ss_thumbnails/sxhpofuorcagxsaulkmt-3-251204082258-7e66bc48-thumbnail.jpg?width=640&height=640&fit=bounds)
![[DSC Europe 25] Yemi Olagbaiye - From Senses to Systems: Building Organisatio...](https://cdn.slidesharecdn.com/ss_thumbnails/zhgiesqoqd68z6lkos3o-5-251203092157-dcce0c99-thumbnail.jpg?width=640&height=640&fit=bounds)
![[DSC Europe 25] Dusan Jovicic - AI Story: From on-prem to cloud and back agai...](https://cdn.slidesharecdn.com/ss_thumbnails/8kp49m6uq22ifnbwhfnk-2-251205085715-964d11a6-thumbnail.jpg?width=640&height=640&fit=bounds)
Introduction to deep learning using python
- 1. Introduction to Deep Learningusing Python Lino Coria [email protected] @linocoria
- 2. About Wiivv Wiivv isa technology company transforming footwear and apparel for every human body. Wiivv Insoles and Sandals are created uniquely for you, based on measurements taken from the award-winning Wiivv app.
- 3. Testing our sandals- Boston Marathon 2018
- 4. What is DeepLearning?
- 5. AI, Machine Learning,Deep Learning Artificial Intelligence Machine Learning Deep Learning
- 6. Artificial Intelligence -Deep Learning
- 7.
- 8. The Power ofDeep Neural Networks Amount of Data Performance Traditional ML Small NN Medium NN Large NN
- 9. Why is DeepLearning in Vogue? ● Hardware ○ GPUs ○ NVIDIA leading the way ● Tons of Data ○ ImageNet dataset: 1.4 million annotated images ● Better Algorithms ● Democratic ○ If you know Python, you can do deep learning ○ Many tutorials, pre-trained models
- 10. Deep Learning andPython
- 11. Open-Source Resources forDeep Learning Keras
- 12. A good wayto start Keras
- 13. How to getData
- 14. Example: MNIST Dataset Aclassic Machine Learning problem Input X (Images) Labels Y: 5 0 4 (Answers)
- 15. Training Data +Testing Data Input X (Images) Labels Y Training Data Input X (Images) Labels Y Testing Data
- 16. Neural Network 1. Input(image pixels) 2. Hidden layers (data transformations) 3. Output (predicted label)
- 17. Weights Layer (data transformation) Input X Weights Layer (datatransformation) Predictions Y’ True targets Y (Labels) Loss function Loss score Optimizer Weight update
- 18. Load MNIST Data >>>import keras >>> from keras.datasets import mnist >>> (train_images, train_labels), (test_images, test_labels) = mnist.load_data() >>> train_images.shape (60000, 28, 28) >>> len(train_labels) 60000 Training input: 60,000 images (28 pixels * 28 pixels) Training labels: 60,000
- 19. MNIST training data >>>import matplotlib.pyplot as plt >>> plt.imshow(train_images[0]) >>> plt.gray() >>> plt.show() >>> train_labels array([5, 0, 4, ..., 5, 6, 8], dtype=uint8)
- 20. MNIST testing data >>>test_images.shape (10000, 28, 28) >>> test_labels array([7, 2, 1, ..., 4, 5, 6], dtype=uint8) Testing input: 10,000 images (28 pixels * 28 pixels) Testing labels: 10,000 Training 60,000 images 60,000 labels Testing 10,000 images 10,000 labels
- 21. Data Reshaping (InputX) # Train images >>> train_images = train_images.reshape((60000, 28 * 28)) >>> train_images = train_images.astype('float32') / 255 # Test images >>> test_images = test_images.reshape((10000, 28 * 28)) >>> test_images = test_images.astype('float32') / 255 1. 28 * 28 image (int values 0 - 255) 2. 28 * 28 = 784-element vector: [row 0, row 1, …, row 27] 3. 784-element vectors with float values in range 0 to 1
- 22. Data Reshaping (LabelsY) >>> from keras.utils import to_categorical # Train labels >>> train_labels = to_categorical(train_labels) # Test labels >>> test_labels = to_categorical(test_labels) 1. train_labels = [5 0 4 ... 5 6 8] 2. train_labels = [[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.] [1. 0. 0. 0. 0. 0. 0. 0. 0. 0.] [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.] … [0. 0. 0. 0. 0. 1. 0. 0. 0. 0.] [0. 0. 0. 0. 0. 0. 1. 0. 0. 0.] [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]]
- 23. Building our Network >>>from keras import models >>> from keras import layers >>> network = models.Sequential() >>> network.add(layers.Dense(512, activation='relu', input_shape=(28 * 28,))) >>> network.add(layers.Dense(512, activation='relu')) >>> network.add(layers.Dense(10, activation='softmax'))
- 24. Neural Network 1. Input(28 * 28 = 784) 2. Hidden layers (data transformations, 512 elements per layer) 3. Output (predicted label: 10-element vector with probabilities)
- 25. Weights Layer (data transformation) Input X Weights Layer (datatransformation) Predictions Y’ True targets Y (Labels) Loss function Loss score Optimizer Weight update
- 26. Network compilation >>> network.compile(optimizer='rmsprop', ...loss='categorical_crossentropy', ... metrics=['accuracy'])
- 27. Weights Layer (data transformation) Input X Weights Layer (datatransformation) Predictions Y’ True targets Y (Labels) Loss function Loss score Optimizer Weight update
- 28. Let’s train! >>> network.fit(train_images,train_labels, epochs=5, batch_size=128) Epoch 1/5 60000/60000 [==============================] - 2s 33us/step - loss: 0.2179 - acc: 0.9329 Epoch 2/5 60000/60000 [==============================] - 2s 31us/step - loss: 0.0805 - acc: 0.9750 Epoch 3/5 60000/60000 [==============================] - 2s 32us/step - loss: 0.0531 - acc: 0.9841 Epoch 4/5 60000/60000 [==============================] - 2s 31us/step - loss: 0.0385 - acc: 0.9883 Epoch 5/5 60000/60000 [==============================] - 2s 31us/step - loss: 0.0290 - acc: 0.9908 Train Accuracy: 99.08%
- 29. Test Accuracy >>> test_loss,test_acc = network.evaluate(test_images, test_labels) 10000/10000 [==============================] - 0s 47us/step >>> print('Test accuracy: %s' % test_acc) Test accuracy: 0.9799 Test Accuracy: 97.99%
- 30. Resources Online Udacity, Coursera PyImageSearch Book Deep Learningwith Python by Francois Chollet
- 31.
- 32.