- Model word problems as computational problems.
- Determine an appropriate algorithm design paradigm for a new problem.
- Design an algorithm for a new problem and argue its correctness.
- Identify and cope with computational problems that are infeasible.
- Consider the implications of modeling decisions in the real world.

- The concepts in an algorithms course often take a little bit of time to click, and are somewhat independent of each other. We care more about whether you understand a topic by the end of the quarter than in the middle. We hope the format will allow you to return to questions and topics you missed the first time.
- We hope to give you
*some*flexibility in which problems you do, allowing you to focus on the aspects of the course that are more important to you.

Grade | Description |
---|---|

Excellent (E) | Main idea and edge cases are all correct. Would have gotten full credit (or extremely close) with points-based grading. |

Satisfactory (S) | Main idea is correct, but some edge cases or follow-up questions are wrong or missing. Would have gotten about 80-90% on points-based grading. |

Not Yet (N) | Some important error is made, but substantial progress toward a solution. Would have gotten above 50% on points-based grading. |

Unassessable (U) | Directions not followed (e.g., used a library or approach that isn’t permitted) or otherwise shows no substantial progress. |

The main goal with this grading system is to change the focus (and time investment) of grading -- instead of spending significant time focusing on differentiating 1 point mistakes from 2 point mistakes, TAs can spend time generating helpful feedback, and then spend some additional time on grading resubmissions.

We will release 8 homeworks during the quarter. The assignments will often have more problems on them than we expect you to answer, allowing you to choose which fits better with your interests.

**Mechanical problems** usually ask you to execute an algorithm or give short answers, like an input that causes bad-behavior. These problems are usually shorter and require a surface-level understanding of what was discussed in lectures.

**Long-form problems** generally involve designing an algorithm -- either by-hand (in pseudocode) or in actual code. Long-form problems also might ask you to consider real-world effects of using algorithms or to prove facts about algorithms. These problems usually require applying the contents of lectures to new scenarios.

For example, if on homework 1, you got Es on the two long-form problems and then submit a new problem from HW1 long-form problem with HW3 that gets an E, you will then have 3 Es on long-form problems on homework 1.

If on homework 1, you got one E and two S on long-form problems and then resubmit a problem from HW1 scored S on the HW3 resubmission box and it gets an E, then your score for homework 1 long-form problems will be 2 E's and 1 S.

If on homework 1, you got one E and two S on long-form problems and then resubmit the problem scored S with HW3 and it gets an N, then your score for homework 1 long-form problems will still be 1 E and 2 S's (i.e., your grade cannot go down from resubmission).

Coding problems will also be submitted to gradescope, with test cases provided. For coding questions, your score is based on:

- the tests in gradescope
- following any problem specific directions (e.g., if we say you cannot use recursion then your final code cannot use recursion).

Score | requirement |
---|---|

E | all tests pass. |

S | at least 80% of the points from tests on gradescope. |

N | at least 50% of the points from tests on gradescope. |

We do not have hidden tests (other than manually verifying the problem-specific directions).

We will verify requirement 2 for all problems at the end of the quarter, after the submission boxes have closed -- the directions should leave no ambiguity as to whether you are following them or not; please ask us during the quarter for clarification if you aren't sure whether you're violating one.

If you wish to get a higher grade on a programming question after the deadline, you can continue to work on that problem (the gradescope tests will remain accessible). **You must still use a resubmission slot to update your score on the problem**. You are encouraged to continue putting new versions of the code up on gradescope (you don't use any resubmission slots by updating code there).

We will have these minimum guarantees when we convert from course submissions to final grades. That is, if you meet the conditions in the table, your final grade can be no lower than the grade shown in that row.

Grade | Mechanical Problems | Long-Form Problems |
---|---|---|

4.0 | 6E and 2S | 22E and 1S |

3.5 | 6E and 1S | 18E and 4S |

3.0 | 5E and 1S | 14E and 3S |

2.5 | 4E and 1S | 8E and 8S |

2.0 | 2E and 2S | 6E and 7S |

0.7 | none | 10S, with at least one on five different HWs. |

To meet the guarantee you must meet every requirement in the list. So, for example, if you never submit a mechanical problem, you could only meet the 0.7 guarantee.

In the chart “extra” E grades count as S grades. So if you have 22 E grades on long form problems, you meet the long-form requirement for a 3.5.

You must meet every requirement to be *guaranteed* of a particular grade, but we may occasionally give a grade to someone who has not met all requirements for a guarantee (see interpolation below).

We will use the full 4.0 grade scale (it is possible to get a 3.7 or 3.3, for example, even though those rows aren't in the table).

At the end of the quarter, we will develop formulas to interpolate between grade breaks, based on these factors

- The number of S or better scores on long-form problems beyond required.
- The number of E scores on long-form problems beyond required.
- The number of S or better scores on long-form problems beyond required.
- The number of E scores on long-form problems beyond required.

We will use these factors to decide on a grade ``between'' the grades shown. For example, if you meet the requirement for a 3.0, but don't meet the requirements for a 3.5, we will ``interpolate'' to decide on a grade between 3.0 and 3.4.

The formulas will place weight on these factors in the order listed (e.g., category 1 above will always have at least as much weight as category 2). We will not release these formulas at the end of the quarter, nor will we debate whether another formula might have made for different results.

The formulas may vary for different ranges (e.g., there may be different weights for the [3.5, 3.9] range than for the [2.5, 2.9] range).

When we interpolate, we may occasionally move a grade at the next guarantee or higher. For example, if you have only 5 E grades for mechanical problems, but you have 24 E grades for long-form problems, you don't have a **guarantee** of a 3.5 from the chart, but you would end up with a grade of 3.5 or above, since you were well above the requirement for long-form problems. These changes are rare and purely at the discretion of the instructor.

N and U scores do not have a direct effect on calculating grades at the end of the quarter. An N grade means you still had something to learn from doing the problem; the goal of the resubmissions is to give you a chance to take advantage of that opportunity when you can. This can sometimes make an N seem 'worthless', but we encourage you not to think about it that way. Before submitting the problem, you'll have spent time thinking about it: eliminating approaches that don't work, and getting intuition about the problem, even if you haven't gotten all the key ideas yet. You will also get feedback to help you narrow in on the fixes you need.

For students near boundaries that affect whether the course is 'passed' (specifically 0.7 [whether credit is awarded to undergrads], 2.0 [whether the course counts for some majors], 2.7 [whether credit is awarded to graduate students], and 3.0 [whether the course counts for some majors]), we do look for instances where a student is close to a boundary but a number of good-faith attempts at problems only gathered N scores. In such cases, we may decide to increase a grade to above the relevant boundary. These changes are purely at instructor discretion (and only for these grade boundaries). If you are concerned about whether you are on track to pass the course, we strongly encourage you to contact Robbie partway through the quarter to discuss your progress.

- Do not take away any notes or screenshots from your discussion.
- Take a 30-minute break
^{1}before writing up your solution**individually**. - Cite the names of all of your collaborators somewhere in your writeup.

- Definitions and terminology can differ significantly (and subtly!) depending on the author. Be careful that other resources are saying what you think they are saying.
- You may not search with the intent of finding a solution to the exact homework problem being asked.
- You may not use commercial tutoring resources like Chegg for the problems we ask, nor post our materials or your answers to those websites.
- You may not publicly post code you have written for this course, even after the course is over.

**How we think about AI systems**
Artificial Intelligence (including programming tools like co-pilot, and large-language models like chat-GPT) are becoming more useful tools in computer science. While it is good to learn how to use these tools, it is important to gain foundational skills that do not rely on tools.
We encourage you to think of AI-systems similarly to how you think about calculators. Calculators can do arithmetic much better than you can, but it is still useful to be able to do arithmetic without a calculator (for example, to be able to check that an answer from a calculator is reasonable, when a question is in an inconvenient form for the calculator, or when a question is so simple you can answer a question faster than you can access a calculator).
Calculators are much better at multiplying one-digit numbers than you are, nonetheless teachers made you memorize and practice with times-tables and multiplying two-digit numbers. Times-tables are a needed skill to be able to do the basic arithmetic that is still useful even with calculators. In the same way, we may ask you questions that AI can solve more quickly than you can---but if you never practice fundamentals, you'll never be able to solve those more complicated problems.

**Specific Policies**

- You may not ask an AI system to 'do your homework for you'
- In particular, you may not put a question we ask (or a modification of it, or a part of it) into a system like chat-gpt. The staff puts significant effort into homework problems that will help you learn by working with them. You will learn more by working on a problem for an hour (even if you get stuck!) than just by reading a solution; putting questions directly into these systems robs you of the chance to learn.
- You may not use a system like co-pilot (or chat-gpt) to generate code. You are expected to type all of your code yourself.
- But you may
- Use an AI system to help you word things (for example, to help you translate a subtle idea from your native language into English, or to make an idea you have more formal). We encourage (but do not require) that you limit yourself to wording one sentence at a time. When using these systems, it is easy to lose track of the main idea, which is what we really want you to learn. Rewording one sentence at a time helps you focus on what's important.
- Use an AI system to generate examples of fundamental operations you've forgotten, or how do do something in a new programming language. For example, questions like 'What does a loop that uses an Iterator for a Queue in Java look like?' is fine
- If you do use an AI system, you should remember:
- You are still responsible for the correctness of your submission. We will not accept regrades that say an AI system mistranslated a correct idea of yours, for example
- You
**must**state that you used an AI system in generating your answer. Cite it as you would a human collaborator, along with generally how you used it (e.g., 'I used chat-gpt to make an explanation I drafted sound more formal in problem 3' or 'I used google translate to translate an explanation I drafted into English'). - You're taking this class for a reason! Whether you're hoping to improve programming skills, prepare for technical interviews, or just learn more about algorithms, there are tasks you'll want to do (soon!) where you might not have access to AI systems at all. You don't want chat-gpt to get you a good grade and then be unable to pass a tech interview because you didn't learn as much as you would have otherwise.

What happened? | Is it a violation? |
---|---|

When searching for general information, you accidentally find the exact question we asked. You tell the staff, and provide a link to what you found. | Not a violation! We’ll say “thanks for letting us know!” and make sure you didn't plagiarize. There won’t be a penalty but only a warm, fuzzy feeling. |

You and a friend separately write up solutions, then compare. Your friend suggests that your conclusion is a little unclear. You formulate a new conclusion on the Zoom call together. | Violation! That is no longer your individual writeup. |

You and a friend separately write up solutions, then compare. Your friend suggests that your algorithm runs in time O(m+n) not O(m^2). You wait 30 minutes, then return to your writeup, decide your friend was right, and update your writeup. | Not a violation! Bug fixes and minor rewordings done by you at another's suggestion are fine. The writeup is still substantially yours. |

You find a textbook with sample solutions to similar problems. You see that they like to introduce variables with “Consider” and use “hence” instead of “because.” You copy these words, because they seem cooler. | Not a violation! Single words or stock phrases are things you can learn from. It is not a violation to emulate style (but "hence" is a little archaic). |

Lectures will be recorded; while in-person attendance is encouraged where possible, so that you can ask questions and discuss with your fellow students, we provide lecture recordings so that if you need to isolate, you can still see lectures (recordings are also available if you wish to review content or just slow material down).

We will have some office hours via zoom.

In the event of illness for the instructor or a TA, we may use any of the following strategies to keep the course running (in order of preference):

- Finding a substitute staff member (to run lecture or conduct the office hour)
- Converting a lecture or office hour from in-person to zoom-only.
- Providing an old recording for lecture or rescheduling an office hour.

Regardless of whether a lecture is conducted in-person or on zoom, we will provide a recording on panopto.