↓Skip to main content

Learning AI Poorly: You can start machine learning in 10 minutes... seriously

·3 mins

(previously posted on LinkedIn)

I finally broke down and replaced my about-to-die 2017 MacBook with a brand new one. For a minute I thought I was going to write something on how easy it is to get a fresh machine doing ML with some follow along steps to do it yourself.

That got real boring real quick so I decided to pivot to something more fun.

Do you have a Google account? Do you have five to ten minutes? Do you want to create a neural network, train it and use it to do some AI? Great! Follow along to do exactly that… for free… on Google Colab.

  1. Go to https://colab.research.google.com/ and log in with your Google Account.
  2. Click on “New Notebook”

  1. Since you’re going to do ML, you want a GPU, right? Great! In the menu, go to Runtime-> Change Runtime Type

  1. Choose “T4 GPU”

  1. Copy the following text:
import tensorflow as tf
tf.test.gpu_device_name()

into the first code cell (see the image to understand what I mean)

  1. Click that little “play” button to the left of the code you just pasted to run it. Something like “/device:GPU:0” will appear just under the code cell. CONGRATULATIONS!!!! You just ran some tensorflow code that checked to see if a GPU was present. That’s some ML codin'!
  2. Let’s actually do some ML. Delete the code you pasted above by highlighting it and hitting delete. Now copy the code below and paste it in the same place.
# Import libraries
from tensorflow.keras.datasets import mnist
import tensorflow as tf

# Load a dataset
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()

# Preprocess the data
train_images = train_images / 255.0
test_images = test_images / 255.0
train_labels = tf.keras.utils.to_categorical(train_labels)
test_labels = tf.keras.utils.to_categorical(test_labels)

# Build a neural network
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')
])

# Compiling the neural network
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

# Train the nerual network
model.fit(train_images, train_labels, epochs=5, batch_size=32)

# Evaluating the neural network
test_loss, test_acc = model.evaluate(test_images, test_labels)

print('Test accuracy:', test_acc)
  1. Hit that same play button. If you scroll down you’ll see some output below your code cell.

Congratulations. You just:

  1. downloaded a dataset of handwritten digits and got it in shape for use by a nerual network.
  2. created a nerual network with 4 layers, Flatten, Dense, Dropout, and Dense. This network will take in a 128x128 pixel image of a handwritten digit and it will output a list of 10 different probabilities. Each one represents the probability that the handwritten digit is a 1, or a 2, or a 3, etc.
  3. trained the network on a series of handwritten images that are labeled with the digit they actually represent.
  4. ran an evaluation step to see how well your network classified each image in a test dataset. In the screenshot above you’ll see the last line says “test accuracy: 0.9771283712987” or something.. that means it is about 97.7% accurate. Hopefully that took less than 10 minutes and now you can say you created your own neural network. neat!
↑