Assignments

For all readings unless otherwise noted, please write down two non “yes or no” questions and type them into the Google doc I’ve shared with you. Be prepared to discuss the questions/readings in class.

Class 10

The Final (due 5/5)

Using what we’ve learned in the class, create a web interface for exploring a dataset. This interface can be portal into a museum collection, a tool that aids in the creative process, or a continuation of your exploration for the midterm.

Metrics for grading:

• 25% - Originality of concept - Does the project explore an original and novel concept?
• 25% - Fit and finish - Is the project finished? Was there clearly thought put into this?
• 25% - Theme - Did the project follow the prompt?
• 25% - Presentation - Is the presentation coherent and thoughtful?

IMPORTANT - email me a proposal for your final by next week, due 4/21

Class 7

The Midterm (due 3/31), proposal due 3/24

Choose an image data set to explore. This can be an existing API or data set such as a digital museum collection, but it could also be a data set of your own creation such as a website that you scrape or collection of your own photos. Using the tools we covered in the last class, create an analysis of the visual topology of this collection. When using machine learning to look with a macro lens, what patterns emerge? What narratives are newly exposed with these techniques?

Be sure to collect at least 500-600 images (1000 or more would be better) in order to have enough to analyze. The final deliverable for this class will be a series of images, along with a quick in-class presentation. If you are comfortable with web development, please feel free to create something larger for the midterm.

Presenting the project:

If there are no group projects then we will have less than 10 minutes per person to present, “pecha kucha” style. I will be expecting a fairly concrete and succinct presentation of about 5 minutes.

Metrics for grading:

• 25% - Originality of concept - Does the project explore an original and novel concept?
• 25% - Fit and finish - Is the project finished? Was there clearly thought put into this?
• 25% - Theme - Did the project follow the prompt?
• 25% - Presentation - Is the presentation coherent and thoughtful?

If you are not interested in creating an exploration like this, you can instead build your own set of classifiers for a custom dataset. Using a multivariate dataset of your choosing, try using the KNN, perceptron, and multilayer perceptron algorithms to create different classification models. Present the models to your class and be prepared to discuss what worked, what didn’t work, and which models you found to work best.

IMPORTANT Due 3/24:

Before class, email me:

• One paragraph describing what your project explore
• If applicable, a link to the data set you will be using
• The name of the data set you will be using

Class 6 (due 3/10)

• Read

  • A New Approach to Understanding How Machines Think - TCAV concept is less important than surrounding context
  • An Overview of Early Vision in InceptionV1

• Code

  • Try using our multilayer perceptron with a different dataset. Can you get it to work with CIFAR-10 or Fashion MNIST datasets? What about the Iris dataset? Try playing with hyperparameters and number of layers/connections. I highly recommend reviewing the referenced material in the class notes if you are at all confused.

Class 5 (due 2/24)

• Read
  • Excavating AI
• Code
  • Continue trying to scrape websites using what we covered in class.

Class 4 (due 2/17)

• Read
  • Using Artificial Intelligence to Augment Human Intelligence - you should have read by now, but we will discuss in class again
  • Creative AI - optional reading, will discuss in class
• Code
  • Review the notes from class and be sure you understand the concept of the perceptron
  • Watch videos/read articles if unclear
  • Try implementing the perceptron on another dataset

Class 3 (due 2/10)

• Read
  • Anatomy of an AI System
  • Don’t Call AI “Magic” by M.C. Elish
• Watch (optional)
  • Vectors, what even are they?
  • Vectors - The Nature of Code
  • Vector Math - The Nature of Code
  • Vector Math II - The Nature of Code
• Code
  • Review the in class jupyter notebook
  • Try implementing your own classifier using one of the other Scikit learn toy datasets. Google around for ideas. This will not be graded or collected.

Class 2 (due 2/3)

• Read
  • Ways to think about machine learning by Benedict Evans
  • The Myth of AI by Jaron Lanier (just read article not comments or video)
• Code
  • Review the in class jupyter notebook
  • If you are having a hard time with Python, try doing all the “Learn the Basics” section at learnpython.org
  • I recommend also picking up Learn Python 3 the Hard Way if you like learning from printed books

Class 1 (due 1/27)

• Read
  • Using Artificial Intelligence to Augment Human Intelligence
• Code
  • Make sure you have Homebrew installed (Mac users), otherwise PC users should have Python3 installed