Max Ehrlich | PlanetTerp

Information	Review
Max Ehrlich CMSC421 Expecting an A Anonymous 05/20/2024	Cons: - tells you not to stress but seems to do everything to make you stress - unclear hw instructions - super hard exam (it was far too long and way too much content) - unclear instructions on final project pros: - lots of curving - no final exam - easy grading on the final project
Max Ehrlich CMSC421 Expecting a B+ Anonymous 05/19/2024	As a student, I was incredibly interested in the possibility of taking an artificial intelligence course at UMD. However the biggest thing that was a let down was the way the course was instructed and the instructors themselves. Max is most definitely incredibly knowledged about the subject, and seems to have a passion for AI itself. However, the course was by far the worst organized 400 level course I have taken at UMD, and that's coming from a university with many "below average" professors in the computer science department. First and foremost, no enforcement of the material was given throughout the year. The course "material" was directly copied and ripped off from CS188, Berkeley's Intro to AI course. The slides, were directly copied as well. As a student, I found it much more effective to visit Berkeley's website and learn from their instructors instead. Signing up for this course would be a waste of your tuition when you can simply learn the material yourself using the CS188 website. Even the midterm itself was given with a low effort on the instructors part. The midterm was given online in class. Almost all students found the midterm to be impossible to complete in the given time, and this resulted in a low average for the exam. Instructors had to remove an entire section of the midterm after grading to better curve the results, which meant that students spent time on that section for no reason. There were 3 total homeworks given in the class. Homeworks 2 and 3 provided more of a learning experience and had more applicability to the course material in comparison to anything else we'd ever learned in class. Homework 3 was building an image classifier using a nueral network which ended up being the most helpful and interesting project done in the class. This project was done using PyTorch, however Max provided absolutely no guidance on how to start with PyTorch leading students to self learn the material before being able to start the homework itself. The final project was a group project, but yet again students were misinformed of the requirements. We were orginally told to complete a short report on our project, but a week before the deadline, this was changed to a 3-5 page report. Moreover, the grading procedure of the final project was not given until a week before the deadline. The lack of communication and help with both the homeworks and the final project speaks to how poorly the class was run. Overall, it seems that some individuals are not meant to teach no matter how knowledgable they are in the material, and Max is one of them.
Max Ehrlich CMSC421 Expecting an A Anonymous 05/10/2024	I'll preface my review by stating that I had Dr. Sujeong Kim, but she co-taught this course with Dr. Max Ehrlich, so much of this feedback applies to them both. Starting with the curriculum, I feel that this course was very poorly designed. The professors put in the bare minimum effort, more or less entirely ripping off Berkeley's CS188 course (https://inst.eecs.berkeley.edu/~cs188/fa23/), including all of the topics and slides. In fact, the slides they provided us with were identical to the Berkeley ones, just with the Berkeley branding replaced by UMD branding. They didn't even bother recording lectures, telling us to use the Berkeley recordings instead. Also, the Berkeley course had regular discussions, homeworks, and projects—which would have helped us learn to apply the material we were learning, something the professors didn’t really seem to care about. Moving on to the quality of the lectures, Dr. Kim seems very knowledgeable, but she simply cannot teach effectively and seems best suited for research. Her lectures pretty much consisted of reading off the Berkeley slides and giving inadequate answers when anyone would ask a question. It seems that there was very minimal preparation done before the lectures. The only reason anyone even bothered attending lectures was the clicker quizzes, which effectively made attendance mandatory. I personally learned more putting in headphones and reading the Berkeley notes instead of listening to her. The assignments didn't make the course any better either… There was one midterm exam, which was a disaster. The professors made us take the exam on Gradescope on our laptops in class, because their goal was not to measure students' comprehension of the material but instead to minimize the time and effort needed to grade the exam while maintaining the façade of a well-run class. They provided hardly any review materials themselves and only posted full solutions after complaints on Piazza. And even the solutions provided turned out to have errors. Besides that, they pointed us to Berkeley's past exams, which ended up being nothing like our exam. The exam was such a time crunch that hardly anyone could even finish, and they had to curve the exam to be out of 78 points even though there were well over 100 points originally. The first 2 homework assignments were quite boring, but Homework 3 (a PyTorch lab which involved building an image classifier from scratch) turned out to be quite interesting. However, the professors did not teach us anything to equip us with the knowledge necessary to do the homework. They did not go over how to use PyTorch, how to design a neural network, how convolutional neural networks work, how to improve models, or anything else that would be remotely useful. While this was one of the only thought-provoking assignments, the professors did not prepare us for it at all. Finally, we had a group project due at the end of the semester that was pretty interesting, but the professors did not communicate the requirements clearly at all. They waited until a week before the end of the semester to tell us the details of the final poster presentation and announce that we had to complete a 3-5 page report in addition to the deliverables that we proposed in our project proposals. And that's also when they finally communicated the grading criteria, which was laughable. Somehow, the poster presentation was worth 70%, while the rest of the deliverables combined (including the 3-5 page report and all of the code) were worth just 30%. Overall, it seems like not much thought was put in this course. Why am I paying for a cheap rip-off of a Berkeley course that is freely available to all?

Max Ehrlich
CMSC421

Expecting an A
Anonymous
05/20/2024

Cons: - tells you not to stress but seems to do everything to make you stress - unclear hw instructions - super hard exam (it was far too long and way too much content) - unclear instructions on final project pros: - lots of curving - no final exam - easy grading on the final project

Max Ehrlich
CMSC421

Expecting a B+
Anonymous
05/19/2024

As a student, I was incredibly interested in the possibility of taking an artificial intelligence course at UMD. However the biggest thing that was a let down was the way the course was instructed and the instructors themselves. Max is most definitely incredibly knowledged about the subject, and seems to have a passion for AI itself. However, the course was by far the worst organized 400 level course I have taken at UMD, and that's coming from a university with many "below average" professors in the computer science department. First and foremost, no enforcement of the material was given throughout the year. The course "material" was directly copied and ripped off from CS188, Berkeley's Intro to AI course. The slides, were directly copied as well. As a student, I found it much more effective to visit Berkeley's website and learn from their instructors instead. Signing up for this course would be a waste of your tuition when you can simply learn the material yourself using the CS188 website. Even the midterm itself was given with a low effort on the instructors part. The midterm was given online in class. Almost all students found the midterm to be impossible to complete in the given time, and this resulted in a low average for the exam. Instructors had to remove an entire section of the midterm after grading to better curve the results, which meant that students spent time on that section for no reason. There were 3 total homeworks given in the class. Homeworks 2 and 3 provided more of a learning experience and had more applicability to the course material in comparison to anything else we'd ever learned in class. Homework 3 was building an image classifier using a nueral network which ended up being the most helpful and interesting project done in the class. This project was done using PyTorch, however Max provided absolutely no guidance on how to start with PyTorch leading students to self learn the material before being able to start the homework itself. The final project was a group project, but yet again students were misinformed of the requirements. We were orginally told to complete a short report on our project, but a week before the deadline, this was changed to a 3-5 page report. Moreover, the grading procedure of the final project was not given until a week before the deadline. The lack of communication and help with both the homeworks and the final project speaks to how poorly the class was run. Overall, it seems that some individuals are not meant to teach no matter how knowledgable they are in the material, and Max is one of them.

Max Ehrlich
CMSC421

Expecting an A
Anonymous
05/10/2024

I'll preface my review by stating that I had Dr. Sujeong Kim, but she co-taught this course with Dr. Max Ehrlich, so much of this feedback applies to them both. Starting with the curriculum, I feel that this course was very poorly designed. The professors put in the bare minimum effort, more or less entirely ripping off Berkeley's CS188 course (https://inst.eecs.berkeley.edu/~cs188/fa23/), including all of the topics and slides. In fact, the slides they provided us with were identical to the Berkeley ones, just with the Berkeley branding replaced by UMD branding. They didn't even bother recording lectures, telling us to use the Berkeley recordings instead. Also, the Berkeley course had regular discussions, homeworks, and projects—which would have helped us learn to apply the material we were learning, something the professors didn’t really seem to care about. Moving on to the quality of the lectures, Dr. Kim seems very knowledgeable, but she simply cannot teach effectively and seems best suited for research. Her lectures pretty much consisted of reading off the Berkeley slides and giving inadequate answers when anyone would ask a question. It seems that there was very minimal preparation done before the lectures. The only reason anyone even bothered attending lectures was the clicker quizzes, which effectively made attendance mandatory. I personally learned more putting in headphones and reading the Berkeley notes instead of listening to her. The assignments didn't make the course any better either… There was one midterm exam, which was a disaster. The professors made us take the exam on Gradescope on our laptops in class, because their goal was not to measure students' comprehension of the material but instead to minimize the time and effort needed to grade the exam while maintaining the façade of a well-run class. They provided hardly any review materials themselves and only posted full solutions after complaints on Piazza. And even the solutions provided turned out to have errors. Besides that, they pointed us to Berkeley's past exams, which ended up being nothing like our exam. The exam was such a time crunch that hardly anyone could even finish, and they had to curve the exam to be out of 78 points even though there were well over 100 points originally. The first 2 homework assignments were quite boring, but Homework 3 (a PyTorch lab which involved building an image classifier from scratch) turned out to be quite interesting. However, the professors did not teach us anything to equip us with the knowledge necessary to do the homework. They did not go over how to use PyTorch, how to design a neural network, how convolutional neural networks work, how to improve models, or anything else that would be remotely useful. While this was one of the only thought-provoking assignments, the professors did not prepare us for it at all. Finally, we had a group project due at the end of the semester that was pretty interesting, but the professors did not communicate the requirements clearly at all. They waited until a week before the end of the semester to tell us the details of the final poster presentation and announce that we had to complete a 3-5 page report in addition to the deliverables that we proposed in our project proposals. And that's also when they finally communicated the grading criteria, which was laughable. Somehow, the poster presentation was worth 70%, while the rest of the deliverables combined (including the 3-5 page report and all of the code) were worth just 30%. Overall, it seems like not much thought was put in this course. Why am I paying for a cheap rip-off of a Berkeley course that is freely available to all?