Lightning AI Studios: Never set up a local environment again →

Log in or create a free Lightning.ai account to track your progress and access additional course materials  

1.2 How Can We Use Machine Learning?

What we covered in this video lecture

In the first video, we discussed the four common application areas of machine learning:

  1. Making predictions (e.g., predicting whether an email is a spam email or not)
  2. Compressing data (e.g., for storing or exploring high-dimensional datasets)
  3. Generating new data (e.g., creating new cat images)
  4. Learning a series of actions (e.g., moving a warehouse robot)

As an alternative categorization, we discussed the three classic categories of machine learning in the second video.

  1. Supervised learning (making predictions from labeled data)
  2. Unsupervised learning (discovering patterns in unlabeled data; data compression & exploration)
  3. Reinforcement learning (learning a series of actions)

Additional content

Below are some additional resources on the different subfields of machine learning (and by extension: deep learning):

Log in or create a free Lightning.ai account to access:

  • Quizzes
  • Completion badges
  • Progress tracking
  • Additional downloadable content
  • Additional AI education resources
  • Notifications when new units are released
  • Free cloud computing credits

Quiz: 1.2 How Can We Use Machine Learning? Part 1

Using machine learning, we can compress a 784-dimensional handwritten image into … (Check all that apply)

We can certainly compress data down to 1 dimension (however, it may come at a large information loss).

Correct. We have seen an example in a lecture video.

Correct. This is analogous to compressing the data to 2D, which we have seen in the lecture video.

Correct. The number of dimensions we compress the data to is arbitrary. However, compressing data to smaller dimensions typically comes at a higher information loss.

Please answer all questions to proceed.

Quiz: 1.2 How Can We Use Machine Learning? Part 2

Generating new data can be considered as a subcategory of …

Incorrect. Generative machine learning can be supervised, but most approaches are unsupervised.

Incorrect. Most approaches to generative learning are unsupervised, but supervised methods exist as well.

Correct. Both supervised and unsupervised methods for generative learning exist. Sometimes, the same unsupervised generative model (e.g., variational autoencoder, not covered in this class) can be converted into a supervised one (e.g., conditional variational autoencoder).

Please answer all questions to proceed.
Watch Video 1

Unit 1.2

Videos
Follow along in a Lightning Studio

DL Fundamentals 1: Intro to ML & DL

Sebastian
Launch Studio →
Questions or Feedback?

Join the Discussion