Welcome to 682!

This class will teach you the core methods and principles behind the incredibly exciting revolution currently happening in the world of machine learning: the renaissance of deep neural networks.

As you probably already know, neural networks are revolutionizing many fields including computer vision, natural language processing, speech recognition, and reinforcement learning. They are also rapidly being adopted by industry for applications from product quality control to autonomous driving and more. I hope you will find this class as exciting as I do!

This will be a very challenging class for many of you. The programming assignments require significant mathematical skills, sophisticated use of the Python programming language (including slicing and broadcasting), and a high level of abstract thinking. The programming assignments can take many long hours of work. Please be prepared for this!

I love to have fun in the class and we will cover a variety of topics related to computer vision. However, there is one topic I need to raise that I really don’t like to talk about. But I feel it is critically important that I do. I will discuss it at length here, partly so that I don’t have to spend so much time in class on it.

We sometimes have a large number of “academic dishonesty” issues. Put more simply, a lot of people cheated on their assignments. Most of the cases involved people looking at on-line copies of assignment solutions, which are easy to find on the web. Many of these people made claims including:

• They didn’t understand that looking at solutions to the homework on-line was considered cheating.
• They knew it wasn’t right but they didn’t think the penalties would be too severe.
• They did look at solutions, but they “didn’t do anything wrong”.

I want to try to make sure these things don’t happen again this semester. When somebody looks at an on-line copy of the solution to an assignment, or they work with another student to complete the assignment (which is not allowed), I find myself in a difficult position. I have a few choices:

• I punish everybody uniformly,
• I let some people off the hook and not others,
• I let everyone off the hook.

I want you to know that all of these are terrible options. I hate punishing people, but I also can’t reasonably let people off the hook—it’s not fair to those students who actually did the assignments.

Why do people cheat? I think that most people would never expect to do it, but then they find themselves in a desperate situation. Maybe they’re taking 3 classes, and they find that there is a problem set due for each class on the same day. When they realize there is no way to finish, they feel that they have no choice.

Of course, the ideal way to deal with this is to avoid this situation in the first place. However, I want to emphasize that there are options much better than cheating, even when you are in such a desperate situation. These include:

• withdrawing from one of your classes, so that you can manage your remaining load. Withdrawing can result in a “W” or simply eliminating the class from your transcript, depending upon what you negotiate with the professor. However, this is much better than an “F” or a record of “academic dishonesty” on your transcript.
• take a “0” on part of the assignment. There are many parts of each assignment, and not doing one part may result in an 80%, but this is not going to cause you to fail the class. Cheating, on the other hand, often leads to failing the entire class.
• Try to get advice and understand your options. The options may not be great, but they are better than cheating.

During lecture, I will make sure you understand the rules about what you can and cannot do while working on homework and projects. If you have questions about this, please consult the lecture notes or email the TAs.

As I said, I don’t enjoy talking about this stuff. But it’s much worse to deal with it after it’s happened. I expect that some of you will find yourselves backed into a corner in this class. I hope that most of you will do the assignments as early as possible, come to office hours for help, and, if you do end up in a bad spot, simply accept the grade that comes with not finishing, which is often not that bad.

With that out of the way, let’s get back to deep neural networks and having fun!

– Prof. Maji based on a note by Prof. Learned-Miller