Suggested learning path for a manager

My interest in taking this course is to gain a greater understanding of ML. Weeks 1 and 2 of Course 1 have been great. I used to develop software, so the code labs have been interesting. But my goal is not to do the programming myself - I want to understand the challenges, limitations, and opportunities. After viewing a number of optional labs, I have come to the end of Week 2, and it requires me to write a program…

Can I skip it? Am I on the right course?

My application is AI/ML applied to industrial reliability: predictive maintenance, reliability analysis, etc.

It depends on your personal goals. A large fraction of your grade in this course is via the programming assignments.

Hello @Jason_Tranter,

Course 1 (C1) and the first 2 weeks in C2 are for us to build some basic models without and with neural networks. Because they are illustrated with examples and some realistic data, you will get to see the level of performance that those models will get you. You will also see for yourself how we could improve the outcome by ML techniques such as features engineering and features normalization. They are technical, and they are full of codes, but immersed in those codes are some experience in performance limitations, and skills to break the limitations. If you get your hands dirty, you position yourself closer to the technical minded people, and I believe being down-to-earth is always advantageous, but only you know your situation well enough to tell what exactly that can bring to you.

In C2 W3, it is a week of discussion in model development cycle that you, your project, and your engineering team would have to go through again and again as a part of the whole project development process. It is an iterative improvement process that a project manager can’t skip. The week also discusses model performance trade-off and more importantly, the additional resources we would need to toss into the soup to make it more tasty. For example, in an “underfitting” situation, you would need a bigger model, which can be translated into that you would need more computational power or time. In an “overfitting” situation, your team might need more data which can mean a higher budget in data collection. Your ML engineers or data scientists will ofcourse know what they need to make things better, but you can understand the rationale behind right here before they have to explain the concepts from the ground up. If you don’t hire your team members for their skills to teach well, then you might want to learn those concepts from somewhere else so that you can explain or defend yourself when you are in a room fighting for more resources for your team.

C2 W4 is about decision tree. Now, it’s just a whole new way of modeling data, but it is also a truth that gradient-based decision tree can more easily outperform neural network in small to medium sized dataset. It can be a reason why you wouldn’t want to push your team towards building a neural network as sometimes people just love neural network as it has a bigger name.

C3 W1 discusses anamoly detection and C3 W2 discusses recommendation algorithms. I believe by their names they already imply opportunities, but it is a question of whether you want to, again, get your hands dirty and be down-to-earth.

I think Tom is right that whether you should take this course depends on your personal goal. If you think you want more than reading “AI” news on the internet, and your goal is to put yourself into the loop of your team’s model development process in a way that you can speak the same language with your team, then I would say this MLS is a good choice for you. The assignment isn’t easy to people with no coding experience, but we pay for what we want to gain ;), and I think that will set you apart from people who don’t want to pay for that time and effort.

Happy new year, and cheers,
Raymond

2 Likes

You might get more benefit out of the overview course:

Although the best way to be a good manager is to also be a good engineer!

2 Likes

Thank you very much for your detailed response. I would love to get my hands dirty as I am still a coder at heart, but I also want to continue to get a higher-level view of things before I dive too deep. As @TMosh said, I need to set my goals appropriately. The course has been very helpful in helping me understand the challenges involved in the various approaches. I will take a look at the course suggested by @KaiStarkk as perhaps it will help me see the big picture before I dive too deep.

Thank you to you all for taking the time to answer me.

1 Like

I almost only talked about what you could get from the MLS for you, your team and your project, but now knowing your goal, I would like to share that I always find it helpful to google “{industry} AI” or “{industry} Machine Learning”, then look for articles and conference talks videos by leading industry player, or perhaps consultancy, or perhaps from the academia. They will give us some ideas of what’s going on out there. Ofcourse, we can’t expect for deeper technical discussion unless the talk/article was by a technical personnel, but with whatever we receive from them, I think those are good starting points to further the search or to talk with a technical friend or to talk with our technical colleagues. This doesn’t sound like a learning path, or it is making learning a part of your daily life. Once you know more concretely (e.g. the names of) what you need to learn to help your career or your business, then google or we might be able to get you more ideas.

Cheers,
Raymond

Thanks Raymond. One of the challenges is that some of the articles I have come across will quickly dive down the rabbit hole they chose - i.e., if they determined that method X was best for their application, they quickly get into a deep discussion that involved that method’s application to the given problem they are trying to solve. That is somewhat helpful, but it would be more helpful if I knew more about method X (or if the author explained why they chose that method and rejected the others). But I have found some very useful information out there.

My hope is to be able to see the “big picture” - “all” the techniques that could possibly be used and why a person would choose one technique over another (the pros and cons of each method, etc.).

I am learning a lot in the course and Andrew is a very good instructor.

I see. But we can’t change what the writer of article wrote, and neither would they consult us before drafting their article. They don’t write what they don’t plan to write. And it is very common people write positive things that add value to what they want to express.

Sometimes we are lucky to come by a very insightful article, and sometimes we might have a friend who can analyze it for us, but if we want it to come from ourselves, then the question is how much effort are we willing to pay to find out, for example,

This question has at least 2 parts: what are all those “other” methods? what are the difference between them? I believe the answer to the first question is to read a lot, and the answer to the second is to experiment a lot. Experiments give you facts under a number of assumptions you have made in carrying out the experiment. Of course, we might also just read others’ experiment results.

You might have difficulty in the “reading a lot” step if you don’t know the language, and by language, I don’t mean English, but Machine Learning. I think every writer has to assume some pre-requisite knowledge when they draft an article, such as in an article for trading, they won’t bother to explain the meaning of “fiat money”, which definition of “volatility” was used, or how a “broker” works even though those concepts were vital to the consequence they are discussing in the article.

Perhaps this is why we want to learn from the foundamentals like what is presented in the MLS. Through the videos and assignments, we learn the language of Machine Learning and we experience the success and pain of it which may ultimately prepare us for synchronizing or even resonating with those writers.

Before I replied to you, I googled for courses for managers. There are not many of them, and the materials appear to be quite similar to the MLS’s, but maybe their lecturers would present them in a different style. At least, I do not find a path that we can skip the foundamentals :wink:

From your posts, I still have no idea about your background, no idea on what you want to achieve, and no idea on what you want to more specialize in. All I know is that you want to be able to grip the big picture, and since it can be anything, I would still suggest you to go through the MLS and the DLS (Deep Learning Specialization) for the foundamentals, and read a lot of news, and discuss the news with expertise of the area. If you are looking for project management or cloud management, then other courses might be needed.

Above is my two cents.

Raymond

I’m a manager, and doing this course to get a broad overview. I’ll dive into the other courses after I think.

Sounds like a plan, @James_Tyack. Welcome to the community and happy learning! :slightly_smiling_face:

Best regards
Christian

1 Like

Thanks Christian, excited to join this group and reinvigorate my interest in and passion for these exciting and impactful topics. The courses are fantastic.

@Jason_Tranter my opinion would be to maybe find a more theoretical ML course other than this cause this specialty course requires one to delve into hands-on and from your point of view, you don’t want that.
I would suggest looking around Coursera for a more theoreitical course.
Anyways enjoy!