Midterm exam

English translation

Week 11

Exam scope and review topics

Run and save Notebooks — run, edit, save cells, and reproduce results
Read arrays and explain them with plots — meaning of rows and columns, selection operations, and axis labels
Collaborate with a change history — Commits, Branches, Merges, and collaboration records
Tensor Shapes and operations — predict shapes, reshape, and Broadcast
Linear Regression and learning — model outputs, Loss, and Parameter updates
Binary classification and probability — distinguish Logits, probabilities, and Class decisions
Multiclass Classification and Logits — Logits, target Indices, and per-Class results
Layer operations in a CNN — shapes through Convolution and Pooling
Training and evaluation in image classification — data splits, evaluation mode, and misclassifications

Exam guidance

Exam format and provided materials

The midterm is a 60-minute code interpretation exam with handwritten answers. A 1-page exam reference sheet is provided so you can look up the necessary syntax and relationships and apply them to the problem conditions.

Fill in key parts of the provided code and interpret intermediate data values and shapes, outputs, or plots. The exam does not require writing a long program from scratch or memorizing entire programs.

Do not run Colab or AI during the exam. Follow the in-class exercise announcements for the exam date, location, and final procedures.

Unless otherwise specified, answers may be left as fractions, radicals, or symbolic expressions. Draw graphs showing the axes, key values, and direction of change. Equivalent expressions and code with equivalent behavior are accepted.