Types of machine learning

Typical learning problems:

• Supervised learning (this course) (Gmail spam filtering)
  • Training a model from input data and its corresponding labels to predict new examples.
    
• Unsupervised learning (Google News)
  • Training a model to find patterns in a dataset, typically an unlabeled dataset.
• Reinforcement learning (AlphaGo)
  • A family of algorithms for finding suitable actions to take in a given situation in order to maximize a reward.
• Recommendation systems (Amazon item recommendation system)
  • Predict the “rating” or “preference” a user would give to an item.

Some typical types of learning problem include:

• Supervised learning (this course) (Gmail spam filtering) which consists of training a model from input data and its corresponding labels to predict new examples.
• Unsupervised learning (Google News) We are not given a target column.
• Reinforcement learning (AlphaGo) is about teaching agents to interact in the real world.
• Recommendation systems (Amazon item recommendation system) fall under the unsupervised paradigm but with a specific focus on predicting ratings or preferences a user would to certain items.

As mentioned before, this particular course focuses on supervised machine learning.

Supervised learning

Let’s talk about some terminology.

In supervised machine learning, we have a set of observations usually denoted with an uppercase X.

We also have a set of corresponding targets usually denoted with a lowercase y.

Our goal is to define a function that relates X to y.

We then use this function to predict the targets of new examples.

Here is an example, where our X contains emoticons of cats and dogs and our Y contains labels associated with these emoticons. We have our learning algorithm which is a classification algorithm in this case. Using this algorithm we learn the mapping between X and y. Once we had this model we applied it to unseen test data to get predictions.

In this particular case, our unseen data contains the emoticons of a cat and a dog. Our predictions are the labels of cat and dog respectively.

from PIL import Image
from src.toy_classifier import classify_image

img = Image.open("static/module1/alpaca.jpg")
img

The example we saw before classifying images, is an example of supervised machine learning.

In our example, our model was trained on a large number of images and their targets.

We can apply this morning on unseen images, in this particular case, our unseen image is an image of an alpaca.

classify_image(img, 3)

Downloading: "https://download.pytorch.org/models/alexnet-owt-7be5be79.pth" to /home/runner/.cache/torch/hub/checkpoints/alexnet-owt-7be5be79.pth

Class, Probability
llama, 0.72
Eskimo dog, husky, 0.06
Norwich terrier, 0.05

We apply our model to this image and we get these predictions and probabilities scores associated with them. Our model here predicted that this is a picture of a llama with 72% confidence.

Unsupervised learning

In unsupervised learning, we are not given targets and are only given observations X.

We apply some clustering algorithms to create a model that finds patterns in our data and groups together similar characteristics from our data.

In this particular example, our X contains emoticons of cats and dogs. We apply our clustering algorithm on this data and as we get images of dogs and cats clustered together in groups.

Machine Learning Libraries

scikit-learn

There are several machine learning libraries available to use but for this course, we will be using the sklearn library, which is a popular (41.6k stars on Github) Machine Learning library for Python.

What we know so far…

• In supervised learning, we are given a set of observations (𝑋) and their corresponding targets 𝑦 and we wish to find a model function that relates 𝑋 to 𝑦.
• In unsupervised learning, we are given a set of observations (𝑋) and we wish to group similar things together in 𝑋.

So far we have seen that in supervised learning we are given a set of observations X and their corresponding targets Y.

We wish to find a corresponding model function that relates X to y.

In unsupervised learning, we are given a set of observations X and we wish to group similar examples together.

Let’s apply what we learned!