Before we start with this course, I would like you to see a short video that we hook by going out of the studio and asking random people in this field, both the students and non-students alike, two very simple questions. The first question was, what is machine learning? And the second question was, if they think linear regression is part of machine learning. Please watch the video and try to answer those questions for yourself and then please memorize it and then you can come back to these questions later, and see whether your understanding of them will change after you take this course. Thank you. Well, it's a part of artificial intelligence. Basically, it's a way that machines are learning. You give them a ton of data and from that data, they're able to gain knowledge and then ultimately use artificial intelligence to come up with another idea or solutions. So, it's really feeding them data and from that, the machines get smarter, they are able to make decisions. Machine learning helps get inside because I think there's a lot of deception on technology, but there needs to be deception on many more reason to be. It's definitely the next big thing with AI. Machine learning is modeling computers to imitate the human brain and evolve to learn problems, to solve situations. One example I guess machine learning would be Google Self-Driving Car. That was completely machine learning and then beginning it was completely dumb, I had no idea what to do. So, what you do is you have the car just do some actions and then tell it if is good action or not, and then through positive feedback just keep on rearranging nuts routers, a little bit one on one until it satisfies your goal and then, from there you just keep on making it more and more complex, until it solves the task, which is important. [inaudible] we have our students do analysis, mostly analysis. I give them a circuit, they kind of tell me what the circuits do. But if I flipped the problem to a design problem, I tell them this is what I want on the output, the number of components are the degrees of freedom in order to get that circuit to perform, takes some doing, they have to guess. I got guess it's something, I don't have enough equations for all you know, so you have to guess and it's that good feeling guess that usually you wrong in the beginning and then slowly over time you start to develop and we start with that initial guess it's gets a little tough. So for me I think, I can with enough pumping of data what works or doesn't work, I could get somehow some software to do a better job of the initial transaction or do the complete job for me, without me creating spreadsheets with every combination and me looking for areas where oh yeah, this is area that sort of works. I could have something else do that for me. Interesting, so for a little bit I know about neural networks and how you have to train them, I guess, it sounds similar, I'm not the mathematician to kind of give you that. But my gut is you're kind of fitting something expected to something that's kind of not right. I would look at it from stock market point of view. So, I have data, it's kind of random, the Random Walk theory. I've seen people looking into this but can I look at a data set sail the cycles here, not from a point of view of people that invest in the stock market because to this the chart is, and then fundamentalism, I probably have the name for them. So, one person looks at the balance sheet, and says, oh this is the stock you need to buy. Another person says, Oh I don't need to know any of that. Just plot the chalk, put some trend lines and things and we could figure out the direction it's going, oh it broke the trend line and to know. That you could probably get into some algorithm and it's kind of like a linear regression of looking for trends, looking for cycles. I'm trying to predict what's going to happen next. So, if I had a guess, is it close? I'd say it's probably in the ball park. No, it isn't because for linear regression is a separate formula to calculate the line of best fit to go through points and what machine learning does instead is that, it tries to evolve let's say a curve that passes them into two points. So, linear regression well it's like a set formulas, a set formula in some textbook somewhere that says, use a calculative line of best fit. If I were to tell a computer to find me the line of best fit through a machine learning, it'll try and calculate it. There would be lots of different things going into that, like it shouldn't be too accurate, should represent the general situation. I guess I'm lost. There's so many like similarities to it, but I mean there's no one silver bullet, right? So, it's like half and half.
