Embeddings

Exercise 11.1

Intro to Sparse Data and Embeddings

Questions:

  1. Task 1: Is a linear model ever preferable to a deep NN model?
  2. Task 2: Does the NN model do better than the linear model?
  3. Task 3: Do embeddings do much good for sentiment analysis tasks?
  4. Tasks 4–5: Name two words that have similar embeddings and explain why that makes sense.
  5. Task 6: Report your best hyper-parameters and their resulting performance.
  6. Optional Discussion: You can skip this section.

Save your answers in lab11_1.txt.

Fairness in Machine Learning

Do these exercises related to human biases.

Exercise 11.2

Intro to ML Fairness

Questions:

  1. Task #1 — What are the biases present in the given dataset?
  2. Task #2 — Assess the potential bias in some other feature besides education level.
  3. Task #3 — Do as written.
  4. Task #4 — Do you find disparities when you look at race rather than gender? If so, which way to they skew?

Save your answers in lab11_2.txt.

Checking in

We will grade your work according to the following criteria:

See the policies page for lab due-dates and times.