MADS

MADS


Visitor count Binder

MADS

Master of Applied Data Science by University of Michigan

Winter 2021 Cohort

Prologue

The MADS Program is a MOOC type graduate program which is designed for students from all over the world who have full-time/part-time jobs and can only take courses virtually by the University of Michigan, Ann Arbor, School of Information. The MADS course was designed and structured by the instructional designers, served by senior professors from the UM, and taught by other instructors. The course curriculum, videos, the choice of textbooks, and reading materials were created and provided by the instructional designers who developed the course. Other UM instructors teach the courses and hold office hours for students who are attending. The majority of the time, the instructors receive feedback from students and bring them to the department. Sometimes, the instructional designer does not have a chance to get feedback.

If you are a current MADS student, please join the Official MADS Slack Group (You need to use UMICH email to sign in).

Evaluation

Evaluation of the Program

Full Score: 🌟🌟🌟🌟🌟
Overall Score of the program: 🌟🌟
Recommendation:🌟☆
Tuition: 🌟🌟🌟🌟🌟; Structure: 🌟☆; Completeness: 🌟☆
Difficulty: 🌟🌟🌟; Instructors: 🌟🌟🌟; TAs: 🌟🌟☆

MADS Course review:

SIADS 501 Being a Data Scientist

  • Medium workload
  • Reading and writing intensive
  • No coding
  • Score: 🌟🌟🌟🌟☆
  • This is a must-have course for the beginner of Data Science which provides a great deal of knowledge and common sense about Data Science in academic and professional aspects. There are plenty of readings, including articles, news, videos, and podcasts, requiring you to read, watch, and reflect on them carefully and profoundly. It requires the student a high level of time management, critical thinking, and problem-solving skills. It builds up the mindset structure of being a Data Scientist, not just coding and coding. I have learned so much from readings and interviews and love this course. You cannot build a house without blueprints. Same as the data science project. It is essential to clearly understand the project we are working on before we start hitting the keyboard. When I started to study Python by myself last year, I mostly liked hard-coding when I was having trouble cleaning data. Now, when I am facing the same problems collecting and cleaning data again, I try to look for a way of coding to fix the problem instead of manually manipulating data, which is meaningless and a waste of time.

SIADS 505 Data Manipulation

  • Medium/heavey worklaod
  • Coding intensive (Python; pandas)
  • Score: 🌟🌟🌟🌟
  • Very general Python data cleaning course. You will use Python for Data Analysis as the textbook and learn some advanced pandas functions during coding assignments. Even though this is not a reading-intensive course, I would like to recommend you read through the book from the beginning to the end if you want to enhance the ability to manipulate data. The lectures only taught some basic knowledge and simple examples as references. Keep in mind that these examples are just the opener. The student should be good at drawing inferences about other cases from one instance. Use slack group and Stack Overflow heavily and smartly. I have learned how to ask a "clear and good" question on Stack Overflow, and it helps you improve critical thinking and self-study skills.

SIADS 502 Math Method for Data Science

  • Light/medium workload
  • Light reading and writing
  • Light coding load (Python; NumPy)
  • Linear Algebra and Statistics
  • Score: ❌ ZERO (A total disaster course, which I've ever taken, with a very shallow academic standard, lower than the one by universities ranking 100+, delivered by an irresponsible instructor)
  • This is a disordered course structure with many mistakes and errors in lectures and materials. The instructor does not know what students' academic mathematical levels are. Since this is a non-math background data science program which means students have diverse math experiences/abilities. It is impossible to have only a one-month math course to satisfy all academic needs for this program. The lectures for profound math background students is too shallow, but for students with limited math background is too tricky. Four assignments are not related to Python coding, and two writing assignments are confused without practical grade criteria or samples. The instructor, Dr. Erin Ware, reads PPTs word by word without any further explaination, examples or exercises, which is unbearable at all.

SIADS 522 Information Visualization I

  • Heavy workload
  • Coding intensive (Python; Altair; Pandas)
  • Score: 🌟🌟🌟🌟
  • This is the Python course about visualization, and it is time-consuming because I never used Altair to create visualization before. It takes time to get familiar with Altair and handle it very skillfully. The professor provides plenty of supplemental resources about creative visualization and videos about how to generate Altair transforms. Same as 505, it is required the student to go through the Altair documentation while doing assignments. Be aware of managing your schedule if you have a full-time/part-time job because it is a time-consuming course, and you might spend few hours to figure out 1 or 2 questions if you have no experience with Altair.

SIADS 521 Visual Exploration of Data

  • Light workload
  • Light reading and No writing
  • Score: 🌟🌟☆
  • Light to medium coding load (Python; Matplotlib; Folium; Pandas; Plotly)
  • This course is about using visualization tools to present data. It depends on how your vis experience and knowledge you have had. There is nothing complicated or time-consuming part of this course.

SIADS 532 Data Mining I

  • Light to medium workload
  • Light reading and no writing
  • Light to medium coding load (Python; Pandas; Numpy)
  • Score: 🌟🌟🌟🌟☆
  • This is an exciting course, and I enjoy it very much. The professor makes the whole course into several small parts and explains every detail's knowledge and critical points. I love the style of the way of teaching. Even though the assignments are split into segments that need some time to get used to, it reduces each assignment's workload. I think the prof is very good at explaining Math.

SIADS 622 Information Visualization II

  • Medium to heavy workload
  • Coding intensive (Python; Altair; Pandas; NLTK)
  • Score: 🌟🌟🌟🌟☆
  • This is a new course offering the fourth month of Win21. Prof Adar puts a lot of time, energy, and enthusiasm into this course, making the course fantastic and well-organized. The lecture, taught by Prof. Adar, providing many kinds of visualization examples in the real world, looks not very related to the assignments, but it gives me a view of how the visualization is. Even though most of those examples are outdated, it's been designed on purpose because it only contains the essential parts of vis, highlighting the critical knowledge and key points to you, without fancy but useless part, which is attracting too much attention. This course is not only about visualization, generating charts, maps, and so on, but also about the pre-processing of text, NLTK. If you have not learned anything about it, it will help you build the system of pre-processing text and represent it by Altair. Again, Prof. Adar is one of my favorite professors so far.

SIADS 511 SQL and Database

  • Coursera course link: Database Design and Basic SQL in PostgreSQL; Cost: $79/month
  • Light workload
  • Light coding load (PostgreSQL)
  • Score: 🌟☆
  • This is a very entry-level SQL course. About five years ago, I took the Excel to MySQL: Analytic Techniques for Business specialization on Coursera. Somehow, I have MySQL experience, but I never use it since I finished the certification. I regard 511 as a refresh course of SQL. However, the content of SQL by the course is too shallow. I have finished the whole course within two weeks at very slow paces. I do not believe this is a graduate-level course but a college freshman-level course. There's an EXACT THE SAME COURSE (remind: this is the single course, not the specialization) on Coursera which only costs US$79 per month, comparing with UM tuition, US$950 per credit (US$1,267 per credit for non-residents). I don't get the reason the instructor doesn't want to create more and deeper concepts and materials on MADS course since we pay much higher tuition (US$79 vs US$950) and have higher expectation than the Coursera one. I love the structure and curriculum of the course and the way of teaching by Prof. Anthony Whyte. I suggest if you take this course as a SQL refresh course, look for extra exercises and projects by yourself instead of relying on this course all.

SIADS 515 Efficient Data Processing

  • Light workload
  • Light coding load (Python)
  • Score: 🌟☆
  • This course is the introduction of the basic of the linux comman-line interface, debugging concepts, basic algorithmic principles. Emphasize on "INTRODUCTION" and "BASIC". The topic of the course is exciting and valuable. I am willing to study more profound than the concepts by the course itself. All four assignments are elementary. The instructors have done most of the programming and coding. As a student, I didn't do much coding by myself. Even though some parts of the assignment require the student to do, there're some examples in the lectures that are almost the same as their answers. I think the most non-sense assignment is the third assignment, which is also the last one. Only four questions are asking to debug codes. The first question only has four lines of code with only one very obvious bug you need to correct. Same as the fourth question. It has three lines of code, and the bug is too easy to figure out. I don't see any effort in this assignment by instructors. This situation happens to many other MADS courses as well. Some instructors and TAs don't prioritize MADS courses and students. They disregard comments, suggestions, and posts by students until you pin themselves and their boss on the Slack channel. Some TA office hours are less than 10 seconds. There's one only 2 seconds. To some TAs, this job is just an easy-money job, and they treat you differently as those regular students at the University of Michigan even though you paid the same tuition.

SIADS 516 Big Data Scalable Data Processing

  • Light to medium workload
  • Medium coding load (Python; Apache Spark)
  • Score: 🌟🌟🌟
  • This course is also an introduction to the Spark data analysis framework of the analysis of Big Data. It gives you an overall view of using MapReduce, Spark RDDs, Spark DF, and Spark SQL to sort, filter, and summarize datasets. Like other 500-level courses, after finishing all videos and assignments, I don't have enough confidence to say that I fully understand Spark data analysis. I believe that there are much more knowledge and concepts that have not been covered in this course. Most of the time, the students tend to guessing the correct output of assignments instead of understanding it.

SIADS 542 Supervised Learning

  • Coursera course link: Applied Data Science with Python Specialization
  • Light to medium workload
  • Medium coding load (Python; Scikit-learn; Pandas; Numpy)
  • Score: 🌟🌟
  • This course is the same course on Coursera, Applied Data Science with Python Specialization, which costs USD79/month and approx. 34 hours to complete. I'm satisfied with the structure of the course and assignments. However, I'm disappointed with how the school indiscriminately imitates the low-tuition specialization courses to the high-tuition graduate program. Again, comparing USD79/month and USD950/credit (one month, in-state) and USD1,267/credit (one month, out-of-state). I would not say I like how they loot their students' money without adding valuable information, knowledge, and skills.

SIADS 543 Unsupervised Learning

  • Medium workload
  • Medium coding load (Python; Scikit-learn; Pandas; Numpy)
  • Score: 🌟🌟🌟
  • This is a decent unsupervised learning course. The assignments are not hard and there's not many students asked questions during the course. The instructor and TAs worked very well. They answered questions timely and held OHs responsibly. However, there's still a lot of information and materials in PCA and feature engineering areas could be taught. PCA is a core section of Machine Learning and there's no single course teaching it.

SIADS 630 Causal Inference

  • Light workload
  • Light coding load (Python; Numpy; Statsmodel)
  • Score: ☆
  • This course talks about Causal Inference which is very useful and interesting. However the course itself is absolutely not well prepared. The instructor never responses any questions, comments and posts directly on the Slack Channel and pushes all of them to TA and OH. The instructor created a form to collect questions. However, he/she only repeated the conception, like 1 plus 1 equals to 2, on the OH and doesn't answer any questions or clarify any concerns. Don't understand why he/she collects questions, but not answer them. The contents of the slides and video are useless, just copy and paste from the book Master Metrics.

SIADS 642 Deep Learning

  • Very light workload
  • NO CODING !!
  • Score: ☆
  • This is a NO CODING and the ONLY deep learning course in MADS program. If you want to learn how to code a deep learning project in Python, please go away and self-study. The quality of the video and content is too much lower than the University of Mich in-person course. The TOTAL length of 642 videos is only 132 minutes which is even shorter than TWO in-person deep learning courses. Typically, the length of ONE in-person deep learning course is about 90 minutes. If you want to learn real things about deep learning, please check out EECS 498 course taught by Justin Johnson. He also taught Stanford University CS 231N course which is an awesome deep learning course. He uploads and shares all course recordings via Google Drive and you can watch them free with your UMICH account. AGAIN, the University of Michigan treats MADS students different from their in-person students and provides much better and higher qualified instructors, professors, and TAs to their in-person students.

SIADS 523 Communicating Data Science Results

  • Very light workload
  • NO CODING
  • Score: ☆
  • This is not a data science related course. The only valuable information provided by the instructor is the Gale Business Demographicsnow. This course introduces several methods and provides some templates of PowerPoint slides. Instead of data science communication, this is a "How to make a PowerPoint for the presentation" course. The only opportunity to communicate with "someone" is the 5-minute presentation without showing off your face, which means the whole presentation only presents the slides and your voice. Moreover, you don't have an opportunity to speak or communicate with a real person. You just recorded the presentation and sent out the recording, which doesn't make any sense as a communication course. The only part of real communication is chatting with classmates and instructors via Slack. The top three communication skills are active listening, sharing feedback, and communication methods. None of them applied to this course. Since the presentation's audience is the instruction team, meaning they are knowledgeable about data science, you do not need to use different skills to adapt yourself to different circumstances. The result/feedback was provided two or three weeks later, so you do not get timely feedback from the audience. The only communication method is talking to your computer like a robot.

SIADS 524 Presenting Uncertainty

  • Medium to heavy workload
  • Medium to heavy coding load (Python; Pandas; Altair)
  • Score: 🌟🌟
  • This is a course to review Altair library which is creating visulizations. The textbook, videos and assignments don't have much relationships. You don't need to watch videos and read textbook to finish assignments. All need is to find out 622 assingments and recall all painful experiences of using Altair. There should be a course teaching other methods and tools to create visualization instead of using Altair only.

SIADS 632 Data Mining II

  • Heavy to very heavy workload
  • Heavy to very heavy coding load (Python; Pandas; Python Class/Object)
  • Score: 🌟🌟🌟🌟
  • This is an extremely time-consuming course. Do not take this course with other courses. I rated it four stars because the auto grade is awful. Sometimes, I fought with auto grade instead of coding a real thing, and there's one time that my coding was entirely correct, but the auto grade gave a "Partial credit; passed some but not all of the tests" error. The instructor and TAs did not explain this situation reasonably and just adjusted grades. Additionally, they did not answer questions timely on the Slack course channel. You HAVE TO ask them privately. This is inconvenient and inefficient because I always check the course channel first and see if someone has already asked my questions. However, this course trained a lot of coding skills.

SIADS 643 Machine Learning Pipeline

  • Light to medium workload
  • Light to medium coding load (Git; Python)
  • Score: 🌟🌟🌟🌟
  • I should take this course at the beginning of the program if possible because it taught very useful knowledge and skills regarding Git and I would manage my GitHub account better. Basically, 643 taught the flow of data within an enterprise and basic functions of and use Git. A lot of "Git push". I would give five stars if they enrich the course. The instrcutor and TA, Michael Hess and Kris Steinhoff, are very helpful and answered questions quickly.