Introduction
I saw this course in a Facebook advertisement in January. I used my first machine-learning package almost 6 years ago. At that moment, I used R to do that. It’s an issue types classification task by a decision tree algorithm. Later, I tried many other machine learning algorithms such as cluster, association, regression and deep learning. But all of these were done by using the packages in R or Python. I think that’s major the reason why I took this course. I’m curious how it works and as a data scientist, I should know more about this.
The course is a course in the MicroMasters Program of Statistics and Data Science provided by MIT. You’ve to finish the following courses to complete the MicroMasters. But you also can get a certain certification by finishing each course. You can choose to pay or not pay for each course. The most difference is the certification I think. For more details, you can check the link. And if you want to save a little fee, you can try Google ‘edx coupon’ or ‘edx discount’.
The learning experience is quite well. It is just like a formal course in the university. So it’s good for you if you miss student life. You can check your learning progress(scores) easily and the due date of each assignment, homework, project and exam. And the best thing, you can get immediate feedback on assignments, homework and projects. As a student, you know. We don’t want to wait for the result.
I think the project part is the most fun of these activities. It is a hands-on activity. You need to submit your code to the online grader and get the result. You can submit a lot of time. It provides try-and-error opportunities for the student. The programming language used in the course is Python.
After finishing the course, there is one suggestion and two things I learned.
Suggestion
If you’re interested in machine learning and really want to get a certification in this course. You better get a score higher than 70 in homework 0; otherwise, you’re going to through your money into the water. I’m not kidding. The course is really tough compared to Coursera, Udemy according to my personal experience.
The good news is that: You can enroll on this course for free first, then decide whether to pay or not pay. The time limit is around one month. In the other word, after the time limit you can’t change your plan to the one with certification. So give it a try.
Learned Things
First, the fundamental of the machine learning algorithm or deep learning algorithm is an optimization problem. Define the loss function which is similar to the objective function. And define the parameters which are similar to actions. Then minimize the loss by finding the best parameters.
Second, my skill in numpy is enhanced now. I can write tidy code with high efficiency by using the matrix operation and the broadcast in numpy. For example, you don’t need to write any ‘for loop’ of this formula.
Summary
It is a nice course and I learned a lot. Under the hood, machine learning is an optimization problem. I’m more confident I’m on the right career path now. Operation research is one of my favourite topics. Later, I’ll enroll the Fundamentals of Statistics of this program also. I think this one can help me to review my statistics concept well.
My certification:
