# Midterm

📗 Coverage: supervised learning W1 to W5.
📗 Number of questions: 40
📗 Length: 2 hours
📗 Sec 1 Midterm A: CS1240, July 10, 12:30 to 2:30 (might start and finish late by 30 minutes)
📗 Sec 1 Midterm B: ES228 (Educational Sciences), July 12, 12:30 to 2:30 (might start and finish late by 30 minutes)
📗 Sec 2 Midterm A: the Beatles, July 11, 5:30 to 7:30
📗 Sec 2 Midterm B: the Beatles, July 16, 5:30 to 7:30
📗 Formula sheet (will be included as a part of the midterm): Link
📗 Sample midterm: (updated July 8) Link
📗 Midterm Version A: Link
📗 Version A Answers: ABEDE ECDDC CCBCC CEDBB CEECD DDDBC DBBAA AAADC
📗 Midterm Version B: Link
📗 Version B Answers: CCABD DAECE BCADC CCEBA DDCCD DDCCA AADBC ABDAB

📗 Questions that will be on the midterm:
M1 File: Q1, Q5 (you will be given eigenvectors)
M2 File: Q2, Q4 (with KNN classifier)
M3 File: Q1, Q4
M4 File: Q2
M5 File: Q3, Q4, Q5
The questions may be slightly changed, see sample midterm.

📗 Each correct answer receives 1 point, each incorrect answer receives -0.25 points, each blank answer receives 0 points. See Link for explanation.
📗 Calculator: NO. (Deduction for bringing a calculator: 2 points.)
📗 Notes: NO. (Deduction for bring an additional page: 2 points. NO examples, quizzes, homeworks questions and answers on the additional pages: deduction for each violation: 2 points.)

# Coverage

📗 Math to do by hand:
(1) Derivatives of composite of polynomials, rational, exponential, and logrithmic functions (no trigonometric functions).
(2) Dot product, matrix multiplication (no matrix inversion).
(3) Find eigenvalues given eigenvectors.
(4) log based 2 for powers of 2.
(5) Convolution with 3 x 3 (or smaller) integer matrices.
(6) Joint, conditional, marginal distributions.
More will be added.

📗 Things on the slides that not on midterm:
L4: Drop Out (formula), Generative Adversarial Network (everything), Autoencoder (everything, but will be on final exam)
L5: Distance between planes (formula)
L6: Decision Tree pruning (formula), Adaptive Boosting (formula)
L7: Finite Difference approximation (formula), Gaussian filter (formula), Laplacian of Gaussian and Difference of Gaussian (formula)
L8: Viola Jones face detection (formula), special Conv Net architectures, e.g. LeNet, GoogLeNet, ResNet (everything)
L9: Zipf's Law (everything), Stationary Distribution and Spectral Decomposition (everything)
L10: Gaussian Navie Bayes (everything), Chow-Liu Algorithm (formula), Tree Augmented Network (formula)

(formula) means you do not need to remember or be able to derive the formulas, but you should still be familiar with the concept.

# Review

📗 Sample midterm: (updated July 8) Link
📗 Video going through sample midterm very quickly: Link
📗 Correction: x^(-2) is not convex on (-infinity, infinity), connecting x = 1 and x = -1, the line is below the function. x^(-2) is convex on (-infinity, 0) and (0, infinity) separately. The domain will be specified on the midterm clearly.

📗 Dandi's review session section 1 notes: Link
📗 Tan's review session section 2 notes: Link