Are you looking for tips and strategies from Data Science experts to tackle Math and Coding?
Manav: Hi everyone, my name is Manav and I’m the Chief Data Science Mentor at International School of AI and Data Science and this is Episode 20 of Data Science and AI Weekly.
If you’ve not subscribed to this podcast, just go to the subscribe button and subscribe immediately so that you don’t miss the next episode when that is going to get released and we release quite a few episodes on a weekly basis.
So in this episode, we’ll continue with our conversation with one of our star faculty at Accredian, Deepesh Wadhwani. So we were having a chat in the previous episode with Deepesh on his journey. Now we are going to take a turn and we are going to talk about getting you prepared for one of the Machine Learning tracks that he does, which is Machine Learning Foundation. So welcome Deepesh again to Episode 20!
Deepesh: Thank you so much for having me here.
Manav: In math itself, let’s say Logistic Regression or Linear Regression or any other model when you talk about all of these Machine Learning models from an intermediate level, you don’t need to learn a lot of maths, but when you start getting into the assumptions then you start seeing that there are a lot of terminologies itself that you have to have to get comfortable with, right?
Then there are various curves that you need to know as you start proceeding further. So at some point, the student needs to know that to what extent does he need to go so that should he focus on that in-depth maths in the beginning, or should he just you know, take things as it is on the foundation level in terms of math and focus on the real-world application and think about getting deeper into mathematical concepts of Data Science at a later point of time.
Deepesh: I personally prefer a student think about the worldly perspective of Data Science more than the math because for math we all are there.
We will be able to help them out with math when it comes to a particular situation. So we’ll be able to help them understand what the question means. The one thing that will be with a student for life will be the logic that he develops.
And that development will happen only if he thinks about how to apply that algorithm in his own business. So let’s say a person is working in it, or a person is in a bank. So an algorithm can be adapted for the domain, the student is in, job well done!
That’s the objective. We all need to get inspired. Everybody else would be using a particular algorithm in some particular context. If we can adapt the same in our domain, we actually created something new in a positive direction.
Manav: Right. So essentially, just to summarize, what you’re saying is that focusing on the real-world application is more important than getting lost in the math jungle itself. Correct. Right. Okay. So that’s the first challenge or that you spoke about.
The second challenge is our students come from various industry domains and they come from experience all the way from they’re Software Engineers, Solution Architects, they are Directors, Vice Presidents and people from programming backgrounds, non-programming backgrounds as well.
Right. So when people are coming from so many different backgrounds, what is one thing that you think everybody needs to keep in mind when they’re learning Machine Learning, because when they start learning Machine Learning, essentially the honeymoon is over because the honeymoon was in the data analysis, part, Machine Learning that brings them to the reality.
Deepesh: So the one thing that I suggest everybody doing, no matter where we come from, no matter if you have 1 year of experience, 10 years of experience, or 25 years of experience, at least during the Machine Learning course, we must do some hands-on until we get in to swim in the data that we have, we will not be able to get that intuitive or friendly feeling from data when a new data point comes in.
So again, I understand somebody who has just 1 year of experience or is a fresher might want to take up a job or somebody who has 25-30 years of experience might want to run a company that does Data Science.
Nonetheless, both of them will need to know how good or bad a particular algorithm is, will that be applicable in my domain or not? Or should I not use Data Science at all? A particular problem can be solved better with heuristics. So, we will not be able to judge this until we solve a few problems at least ourselves.
Manav: Fantastic! And this is what even I emphasize a lot that what is important is that these are what you call the core skill. So, no matter the application the core skills are something that you need to master; no matter whether you want to become the Head of a Data Science practice or whether you want to become a Junior Data Scientist.
So now let’s talk about programming. Because programming a lot of students that we get; 40% of the students are from the non-programming background, they are from IT backgrounds, but they are not possibly working on programming on a day-to-day basis. So what level of programming is required in Machine Learning and if you can give some instances also through some of the libraries then that would help newbies understand that what is the actual level?
If you can give a more realistic picture, that would be great.
Deepesh: But before I answer that question, let me take an example here. Have you used a scientific calculator before?
Manav: Yes.
Deepesh: Everything on the scientific calculator you used?
Manav: Absolutely not!
Deepesh: No, right. So that’s what Python is. Python does everything left, right and center, We are using basic calculations out of it.
So the first thing which I tell during the start of the Machine Learning track is even if you know a lot of programming, it’s not a lot of benefit here because we will be using very few commands, we will be using libraries which are already pre-built.
So everything that we study, we use, let’s say spend 2 hours in a particular session. So everything we study in 2 hours is part of just 3 lines of code. So all we are trying to do is see the logic of how this code works in the background.
And yes, indeed. This should be done. So you give the few names; the libraries; Pandas, Sk Learn, NumPy, Scipy, all of these libraries do basic computations, they build one and another. And ultimately, Sk Learn-it created the complete algorithm in the background, all we need to do is write a 3 liner code to run the algorithm. So I am a Mechanical Engineer, I used to hate programming, I still got into in Data Science, it’s actually fun!
Manav: Fantastic!
So I think you will be a lot of inspiration to a lot of Mechanical Engineers, Civil Engineers, Electrical Engineers, and possibly for people like me as well.
I come from an Electronics Engineering background, and I concur. They’re totally with you that programming is nothing but logic, right? And especially Python as a programming language. The way I say it, it’s simple logic, right? So Python is helping you do things in a simple way.
Right now, we spoke about maths, we spoke about programming. Now, one of the questions I have is that when people are undergoing a Data Science program and this is going to be a wrap-up question before we do the next episode, from a programming point of view, how much practice is required beyond the classes that they are attending on Saturdays and Sundays; on the weekends with you; during the week, right? And what schedule would you recommend to people from to master programming?
Deepesh: So this goes for both Programming and Machine Learning as we move forward in the curriculum.
But let’s say we have a four-hour session on the weekend. I recommend four hours during the week as well. This does not is just programming, have a revision session, make notes during sessions, read those through.
Think about where you can apply in your own business domain.
And of course, we have a programming demonstration in the class, try to replicate that on some other data set. We have a very beautiful collection of data sets and from domain ranging from banking to mechanical engineering to health-care to even the one which we can use on neural networks!
So endless data sets! All of those are available with us. It’s available online, there are a lot of repositories, we can choose data set from our own domain and practice that will take roughly half-an-hour, 45 minutes, maybe an hour in the starting, as we move forward as we realize that Data Science is very structured, it’s like Tetris, we have to just arrange the right block at the right position.
And those blocks we already have pre-built. We have those in our Jupyter notebooks. All we need to do is make sure the right block goes in the right sequence.
Manav: Fantastic! So that was Deepesh Wadhwani, star faculty at Accredian.
And what Deepesh has to say is that if you are motivated, if you are asking the right questions, and if you’re using logic, programming should not be a hindrance, and possibly he himself epitomizes it at the most because he is a Mechanical Engineer himself.
So this is Episode 20 of Data Science and AI Weekly. I am in chat with Deepesh Wadhwani and we will be back with another episode; Episode 21st of Data Science and AI Weekly in which I’m going to continue my conversation with Deepesh and ask him a few more questions that I would personally want to ask him on your behalf.
Thank you very much for tuning in and see you in Episode 21. Thanks for tuning in. This is Manav, I’m signing off!
1 comment