Get best of the week

Create a Powerful Machine Learning Course: Mission Complete



We had 2 unsuccessful starts, 169 students, 8 angry reviews, 3 name changes, too much theory and not enough real practice. Not that it was a complete failure, but if you started teaching Data Science, you need to do it perfectly. Today you will hear a story about how we developed in OTUS the direction of data analysis and which ones we made on this path, and then we corrected the errors.


The first com - pancake


Three years ago, we launched our first data analysis course and called it “BigData Developer” . It was designed for 128 ac. hours, half of which are webinars, and the second is homework and a project. Machine learning and neural networks have ceased to be the lot of the elite and have become a necessary business tool for effective development. IT corporations, online stores, marketing agencies, start-ups and digital services are lined up for the date by Scientists. Jobs flew. The labor market especially acutely felt the lack of Middle and Senior specialists.

It was necessary to teach and grow middle, but no one knew how to do it well. We invited a teacher, developed a program, and as a result, a course came out that, on the one hand, was difficult and even too demanding on the level of training, and on the other, not practical enough.

Confused students with a name change


When we collected the feedback of the first launches, we found that the name does not accurately reflect the essence of the program . Under the term of one section, we taught all the tools of Data Science. In 2018, we re-launched a course called Data Scientist, implying that he is preparing for this profession. After processing it, the volume of webinars increased by 10 hours, but the practice was still a weak point. Most of the tasks were toy examples, far from real tasks with real data sets.
This time the reviews were controversial. Some scolded the course for superficial knowledge, others said that it turned out to be too difficult, although they successfully passed the entrance test for it. Some thought that by changing the name we tried to hush up the not-so-successful first launches. In addition, force majeure happened at one point: the course lost the leader, and then the producer.

Inspiration and a new teacher


Meet Dmitry Sergeyev, author and leader of the Machine Learning course. Together with him, a complete rethinking of the direction of Data Science came to OTUS. We abandoned the idea of ​​putting all practices into one course and made in-depth programs separately for Machine Learning and Neural Networks in Python.

Dima has been analyzing data since 2012. He enthusiastically approached the development of classes for OTUS, filling them with practical chips and interesting tasks.

Key course differences

"Big Data Developer""Data Scientist"Machine Learning
Year2017- beginning of 201820182019 -...
TitleReflects 1 tool, although in fact the course was on different Data Science toolsThe course was not practical enough and detailed, so that having passed it you could consider yourself a serious specialistNow this is one of a series of courses in the Data Science section. The name reflects the essence - the course is dedicated to advanced machine learning practices and only partially affects neural networks
The amount of hours128138178
Webinar hours647470
Watch for independent work6464108
Number of practical exercisessixteen12nineteen

We asked Dima himself to tell how and why he revised the course.

OTUS: Dima, you saw the previous program. How has she changed in the new course?
— , . , , , . - . .

, — «» «». , .

. — . , , , , , .. , .
OTUS: , .
— . , , ( !) . , — - , , .

, , , «» , , — , . , — . , , .

, — . , - , .

. , , , Moons.

, , .
OTUS: , , — ?
— , . 5 , . , .

. , , . .
OTUS: ? ?
— -, . , , , ARIMA, . , — tf-idf .

-, , . , , ( ), «» . , , , , , , AWS ..

, , — , , , , , . , , .

— , . , , , . , , .
«Machine Learning» 28 . , . .

, 24 20:00 . . !

More articles:


All Articles