这是一篇来自美国的关于对机器学习领域的介绍，包括分类、回归、降维、聚类和模型选择的基本技术代写

Workload:

This is a 3-credit course. The median student should spend ~3 hours/week/credit → 9 hours/week for a 3 credit course. A typical week should include:

• a “chalkboard lecture” in person meeting about 2-2.5 hours long
• a “Colab notebook” video about 1-2 hours long
• a homework problem set (estimated time: 2-3 hours)
• a lab assignment (estimated time: 2-3 hours)

A student who is not very comfortable with the prerequisites may have to spend more time than this estimate in order to do well.

Grading:

Homework problem sets: 25%

Lab assignments: 25%

Midterm exam: 25%

Final exam: 25%

The two lowest grades in each of the “Homework problem sets” and “Lab assignment” categories will be weighted 50%. For example: if you earn [50, 75, 80, 90, 64] on labs, your weighted lab average is [75+80+90+(50+64)/2]/4 = 75.5.

Problem sets:

The goal of the homework problem sets is for you to practice remembering and using concepts from the lecture + Colab lessons.

Homework problem sets will be distributed on PrairieLearn, a platform that (1) immediately grades your submission for each question,(2) tells you whether or not your answer was correct (sometimes with extra feedback and comments to help you understand your mistakes),(3) and allows you to re-attempt with a new version of the question.

You can repeat homework problems as many times as you want (including after the deadline, although you won’t earn credit then).

You will never lose credit from extra practice – your overall score on a homework assignment will never go down.

Because the problem sets are auto-graded, they are helpful for meta-cognition – the practice of reflecting on your own thinking and learning, and planning your learning methodology. When you get instant feedback on your work, you can identify areas of weakness, and use this to guide your learning efforts (for example: review the relevant part of the lesson, ask about your misunderstanding on Ed or in office hours, repeat the practice problem in future weeks, etc.)

Lab assignments:

The goal of the lab assignments is for you to go beyond the basic models and techniques discussed in class, and work on much more challenging problems – especially those that require a deeper understanding of the course content.

Lab assignments will be distributed as Python notebooks, with some parts missing. You will work through the notebook and fill in the missing parts. Then, you’ll answer questions about it in PrairieLearn.

You’ll also submit your completed notebook.

Late work policy:

The deadline for each assignment will be listed in NYU Brightspace. You are required to submit your work in PrairieLearn before the deadline.

Make sure to give yourself plenty of time, so that in case you have questions or need help, you have time to ask and get support on Ed well before the assignment is due.

I understand that occasionally, life circumstances may prevent you from fully completing an assignment before the deadline. There are many homework assignments over the course of the semester, so each one individually carries relatively little weight. There is an added element of “forgiveness,” in the two lowest grades counting less. These elements of the course are designed so that the impact on your overall course grade will be small if you miss one assignment or part of an assignment.

However – late work will not be graded for credit unless recommended by the student advocate. In case of illness or other exceptional situations, please contact the student advocate as described in the “Illness or other exceptional situation” section of this document.

Note that in case of an “illness or other exceptional situation” for which the student advocate recommends an extension, you are still required to submit the work as early as possible – e.g. shortly after the “exceptional situation” is resolved.

Exams:

We will have a midterm exam and a final exam. Both exams will be held in person, on a computer – you will have to bring your own (fully charged) laptop computer to the exam session.

During the exam, you are permitted to use:

• a 4-page hard copy reference sheet that you have prepared,with whatever notes you think you will find useful.

• an online calculator, like Wolfram Alpha or equivalent,
• your own (offline) Python terminal for computation.
• the official scikit-learn API reference pages
• the official numpy API reference pages
• the official pandas API reference pages
• the official Keras API reference pages
• the official Tensorflow API reference pages
• Google Translate (i.e. if you do not understand some question text or instructions during the exam, you may translate it.)

You may not use any other materials or resources during the exam. You may not use the course handouts or notebooks (except for parts you have copied into your 4-page PDF).

Each exam will have the following four sections.

1. Basics of ML models (including a small coding problem)
2. Application to the “real world”
3. Miscellaneous short questions not requiring computation
4. A substantial coding problem

Many of the questions on the exam will be similar (conceptually) but not identical to homework questions you have already seen (especially the basic computation questions in Section 1). Some of the questions (especially in Section 3, the short question section) will be about concepts or intuition in the lecture and Colab lessons that you haven’t necessarily used in the homework assignments.

A practice exam will be released a few days before each exam. This will help you understand the exam format and content. The practice exam will have a time limit, like the real exam, and you should try and solve the practice exam in an exam-like setting (e.g. use only the resources that are allowed on the real exam.)