Parijat Dube

Egomaniacal and would give a bad grade if you push back against tyrannical bullwork in useless topics, or are creative, even if your ideas are groundbreaking

80/100 1 Columbia University: Arts & Sciences Fall 2024 Course: STATGR5242_001_2024_3-ADVANCEDMACHINELEARNING : STATGR5242_001_2024_3 - ADVANCED MACHINE LEARNING Instructor: Samory Kpotufe

Workload is above average (compared to other CS classes). Quizzes are difficult. On average assignments take 8-12 hours of time. Grading is lenient.

You cannot trust even a single word of this instructor. Unless you get it written, Parajit's going to break every single agreement. He is given to harassment of students, especially when his illogical demands receive the inevitable backlash. Just look at the NYU ratemyprofessors ratings of this uncouth abuser. He even threatened to fail hardworking students with unexpected hardships, this madman needs to go. He never teaches the full breadth of what is useful in the industry and instead demands extra work on pointless topics... If you take his course, you will regret your decision to study any topic and this is not recommended, he would ruin your uni life.

How LLM's relate to overall machine learning/nn pipeline. How to fine-tune and design LLMs.

Professor Samory is GREAT! He is soooooooooo nice, and you can tell he cares about your future, he wants everybody to get something from this course, and he wants everyone to have a job with the thing he taught. I would definitely recommend this excellent course to everyone I know if Professor Samory is the only one teaching this course, however, that's not the case. - / 1 Columbia University: Arts & Sciences Fall 2023 Course: STATGR5242_001_2023_3-ADVANCEDMACHINELEARNING : STATGR5242_001_2023_3 - ADVANCED MACHINE LEARNING Instructor: Samory Kpotufe

The content of the course is great. The assignments, though difficult, are designed keeping practicality in mind. You are expected to write a lot of code in the assignments. Professor Dube is very approachable and friendly and will go above and beyond to help students. That being said, the professor is mostly unenthusiastic and lacks energy while teaching the class. The class was late evening and it was a 2.5-hour long class which made it even more sleepy. If you are someone who can attend all lectures and keep up with the course, it is a great course to take. After the course, you'll be pretty well-versed in implementing large-scale deep-learning architectures. The grading is very lenient.

Machine Learning, Deep Learning, Transformer Models, Prompting Strategies/Prompt Engineering, RAG, Fine-Tuning, LLM Benchmarking, and Tool-Assisted LLM

Introduction to LLMs, prompt engineering, fine-tuning, RLHF, quantization, pre-training, agents. I think it touched very well on most hot topics. Definitely industry relevant. Instructors gave their best but I think the class was not fully engaged.

如果有志于做generative ai以及llm的话，这门课绝对是堪称私人订制版的合适，甚至我觉得在上完这门课后也可以继续拿着材料继续深挖，毕竟这本身是cs系的课程。但是对于 那些对这个方向没兴趣的学生，这门课可以说是可有可无的水课了。