Hi all,
Lecture 5: Neural Networks
<https://cs50.harvard.edu/summer/ai/2020/lectures/5/> is now available on
the course website. In the penultimate unit of the course, we continue our
exploration into machine learning with a look at neural networks, one of
the most popular and versatile tools in the modern machine learning
toolkit, inspired by the human brain. We'll look at a variety of different
neural network architectures — including deep networks, convolutional
networks, and recurrent networks — and explore how to build and train
neural networks in Python.
In this project, you'll build a neural network to classify real-world
images of traffic signs. You also have the option of alternatively
completing an *Exploratory Project*
<https://cs50.harvard.edu/summer/ai/2020/projects/5/#exploratory-project>
if you'd like to apply neural networks to a task of your own choosing.
A few other important announcements for this week:
- Quiz 5 <https://cs50.harvard.edu/summer/ai/2020/quizzes/5/> is now
available and is due by 11:59pm ET on Wed 7/29. As a reminder, each quiz is
open-book: you may use any and all non-human resources during a quiz, but
the only humans to whom you may turn for help or from whom you may receive
help are the course’s heads.
- Project 5 <https://cs50.harvard.edu/summer/ai/2020/projects/5/> is now
available and is due by 11:59pm ET on Sun 8/2.
- Sections <https://cs50.harvard.edu/summer/ai/2020/sections/> this
week will be an opportunity to explore neural networks more, see additional
examples of lecture material, and ask questions about this week's concepts.
You're encouraged to attend if you can!
As always, feel free to reach out to me or any of the course staff with any
questions!
All the best,
Brian
Hi all,
Lecture 4: Learning <https://cs50.harvard.edu/summer/ai/2020/lectures/4/> is
now available on the course website. In this part of the course, we
introduce the domain of machine learning, looking at techniques that allow
our AI to learn how to perform a task, without explicit instructions for
how to do so. This week, we'll introduce three broad categories of machine
learning: supervised learning, reinforcement learning, and unsupervised
learning.
In the first part of this project, you'll build a classifier to predict
user behavior on an online shopping website, taking advantage of supervised
learning techniques in scikit-learn <https://scikit-learn.org/>. In the
second part of the project, you'll use reinforcement learning to design an
AI to teach itself how to win at Nim <https://en.wikipedia.org/wiki/Nim>:
through repeatedly playing games against itself, the AI will over time
learn which moves are better than others.
A few other important announcements for this week:
- Quiz 4 <https://cs50.harvard.edu/summer/ai/2020/quizzes/4/> is now
available and is due by 11:59pm ET on Wed 7/22. As a reminder, each quiz is
open-book: you may use any and all non-human resources during a quiz, but
the only humans to whom you may turn for help or from whom you may receive
help are the course’s heads. And remember to submit your quiz via
Gradescope as well!
- Project 4 <https://cs50.harvard.edu/summer/ai/2020/projects/4/> is now
available as well and is due by 11:59pm ET on Sun 7/26.
- Sections <https://cs50.harvard.edu/summer/ai/2020/sections/> this week
will be an opportunity to talk more about machine learning. You're
encouraged to attend if you can!
As always, feel free to reach out to me or any of the course staff with any
questions!
All the best,
Brian
Hi all,
Lecture 4: Learning <https://cs50.harvard.edu/summer/ai/2020/lectures/4/> is
now available on the course website. In this part of the course, we
introduce the domain of machine learning, looking at techniques that allow
our AI to learn how to perform a task, without explicit instructions for
how to do so. This week, we'll introduce three broad categories of machine
learning: supervised learning, reinforcement learning, and unsupervised
learning.
In the first part of this project, you'll build a classifier to predict
user behavior on an online shopping website, taking advantage of supervised
learning techniques in scikit-learn <https://scikit-learn.org/>. In the
second part of the project, you'll use reinforcement learning to design an
AI to teach itself how to win at Nim <https://en.wikipedia.org/wiki/Nim>:
through repeatedly playing games against itself, the AI will over time
learn which moves are better than others.
A few other important announcements for this week:
- Quiz 4 <https://cs50.harvard.edu/summer/ai/2020/quizzes/4/> is now
available and is due by 11:59pm ET on Wed 7/22. As a reminder, each quiz is
open-book: you may use any and all non-human resources during a quiz, but
the only humans to whom you may turn for help or from whom you may receive
help are the course’s heads. And remember to submit your quiz via
Gradescope as well!
- Project 4 <https://cs50.harvard.edu/summer/ai/2020/projects/4/> is now
available as well and is due by 11:59pm ET on Sun 7/26.
- Sections <https://cs50.harvard.edu/summer/ai/2020/sections/> this week
will be an opportunity to talk more about machine learning. You're
encouraged to attend if you can!
As always, feel free to reach out to me or any of the course staff with any
questions!
All the best,
Brian
Hi all,
Lecture 3: Optimization
<https://cs50.harvard.edu/summer/ai/2020/lectures/3/> is now available on
the course website. In this part of the course, we explore how AI can
search for solutions that optimize for some goal — minimizing cost,
maximizing rewards or benefits, or satisfying some constraints as best as
possible, for example. In doing so, we'll explore three different
approaches used in AI: local search, linear programming, and constraint
satisfaction algorithms.
A few other important announcements for this week:
- Quiz 3 <https://cs50.harvard.edu/summer/ai/2020/quizzes/3/> is now
available and is due by 11:59pm ET on Wed 7/15. As a reminder, each quiz is
open-book: you may use any and all non-human resources during a quiz, but
the only humans to whom you may turn for help or from whom you may receive
help are the course’s heads. And remember to submit your quiz via
Gradescope as well!
- Project 3 <https://cs50.harvard.edu/summer/ai/2020/projects/3/> is now
available as well and is due by 11:59pm ET on Sun 7/19. This week, you'll
build an AI to generate crossword puzzles using constraint satisfaction
techniques.
- Sections <https://cs50.harvard.edu/summer/ai/2020/sections/> this week
will be an opportunity to talk more about these optimization problems, see
additional examples of lecture material, and ask questions about this
week's concepts. You're encouraged to attend if you can!
As always, feel free to reach out to me or any of the course staff with any
questions!
All the best,
Brian
Hi all,
Lecture 2: Uncertainty <https://cs50.harvard.edu/summer/ai/2020/lectures/2/> is
now available on the course website. While we've so far focused on
information our AI can know for sure (a game's optimal move, a mine's
location, etc.), we'll now transition to how AI can model uncertain events.
We'll start by looking at the mathematics of probability theory. Then,
we'll explore models — including Bayesian networks and Markov chains — that
AI can use to deal with uncertainty.
For your project this week, you'll have two choices (you only need to
choose one): you can either use a Markov chain approach to implementing
Google's PageRank algorithm for computing the importance of web pages, or
you can use a Bayesian network approach to see how AI can be used to trace
the inheritance of genetic traits through generations.
A few other importance announcements for this week:
- Quiz 2 <https://cs50.harvard.edu/summer/ai/2020/quizzes/2/> is now
available and is due by 11:59pm ET on Wed 7/8. As a reminder, each quiz is
open-book: you may use any and all non-human resources during a quiz, but
the only humans to whom you may turn for help or from whom you may receive
help are the course’s heads.
- Project 2 <https://cs50.harvard.edu/summer/ai/2020/projects/2/> is now
available as well and is due by 11:59pm ET on Sun 7/12. For this project,
you can choose to complete either *Pagerank *or *Heredity*.
- Sections <https://cs50.harvard.edu/summer/ai/2020/sections/> this week
are an opportunity to talk about probability and uncertainty, and ask
questions about this week's concepts. You're highly encouraged to attend!
As always, feel free to reach out to me or any of the course staff with any
questions!
All the best,
Brian
