Share a status update about your project.
Your update should include two parts:
What are you trying to do? Where are you on the journey? Make this as a video of a presentation with two slides (one for each part).
We will discuss the effects of algorithmic recommendation systems on people. As usual, you should post on Moodle and be prepared to raise aspects of your own or others' posts during our in-class discussion.
Technology: ___% Fill in the following list (specific instructions are below the list):
Reading: [__%] comments
Portfolio Fundamentals: [__%] comments
010 (sklearn regression): __
Let’s get the project direction a bit more settled, though it doesn’t have to be set in stone yet. In particular, you’ll need to decide on which of your proposed projects to focus on and who you’re working with.
The following template is provided in your Portfolio repositories under narrative/lab04-nn-regression.ipynb.
Common gotchas on this lab:
Remember to instantiate your modules (e.g., nn.MSELoss()), and use PyTorch losses, not mean_squared_error from sklearn.
This assignment should be turned in on Moodle. Note: released late, so due date postponed a bit.
This reflection is similar in format to Reflection 5 but more flexible. You are, as always, welcome to make a compelling, argument for computing grades for yourself differently from this suggestion.
Facial Recognition Data Facial recognition technologies pose complex ethical and technical challenges. Neglecting to unpack this complexity- to measure it, analyze it and then articulate it to others -is a disservice to those, including ourselves, who are most impacted by its careless deployment.
This assignment should be turned in on Moodle.
This reflection is like Reflection 4, but focuses on what you’ve done since then.
You are, as always, welcome to make a compelling, argument for computing grades for yourself differently from this suggestion.
MNIST Train two different kinds of classifier on all 10 digits of MNIST:
Use the high-level API of chapters 1 and 2 to train a ResNet. Use the low-level approach of chapter 4 to train a one-layer MLP; use a 30-dimensional hidden layer with ReLU activations.
Here’s a video walkthrough!
The following template is provided in your Portfolio repositories under narrative/lab03-sgd.ipynb
Lab 03: Estimate proportions using SGD Task: Debug some code to use stochastic gradient descent to estimate two proportions.