1 00:00:02,130 --> 00:00:04,310 Everyone and welcome back to this class. 2 00:00:08,310 --> 00:00:14,070 This lecture is designed to answer some common questions I get about how difficult this course is and 3 00:00:14,070 --> 00:00:19,260 where it fits in terms of academia, the professional world and its practical applicability. 4 00:00:20,040 --> 00:00:25,500 So rather than answering this question individually, over and over again, it seems like a better idea 5 00:00:25,500 --> 00:00:28,620 to make a lecture about it and answer everybody at the same time. 6 00:00:29,490 --> 00:00:33,930 A lot of people will become confused because this is their first machine learning course, and they 7 00:00:33,930 --> 00:00:36,840 just don't know where it fits in the grand scheme of things. 8 00:00:37,350 --> 00:00:39,780 So is this course academic or practical? 9 00:00:40,110 --> 00:00:41,850 And is it for beginners or experts? 10 00:00:42,120 --> 00:00:43,920 Is it fast paced or slow paced? 11 00:00:49,050 --> 00:00:53,910 I'm going to answer the second question first, is this course for beginners or experts? 12 00:00:55,160 --> 00:00:59,450 Interestingly, the problem with this question is a problem for machine learning, too. 13 00:01:00,290 --> 00:01:03,140 Let me explain why this question is itself problematic. 14 00:01:03,910 --> 00:01:09,830 One of the challenges in natural language processing is that of ambiguity when someone says some words. 15 00:01:09,850 --> 00:01:15,190 The problem is those words can mean multiple things, and it just so happens that I have some courses 16 00:01:15,190 --> 00:01:15,760 on an LP. 17 00:01:15,760 --> 00:01:18,010 So this is a problem I think about often. 18 00:01:18,550 --> 00:01:19,960 So how does that apply here? 19 00:01:20,890 --> 00:01:26,110 Well, when we use the words beginner and expert, we have to ask a beginner from whose perspective. 20 00:01:26,350 --> 00:01:31,900 An expert from whose perspective for some of you, you might think that a beginner is a random person. 21 00:01:31,900 --> 00:01:35,650 You take off the street who knows nothing, but even that is ambiguous. 22 00:01:36,550 --> 00:01:41,080 There are people out there who do not even really know how to use a computer beyond turning it on and 23 00:01:41,080 --> 00:01:42,760 checking their email and browsing the web. 24 00:01:43,510 --> 00:01:48,310 There are people out there who are pros at Microsoft Office, so they know how to use a computer and 25 00:01:48,310 --> 00:01:53,080 make a Microsoft Word document or a spreadsheet, but they've never done any programming before. 26 00:01:55,020 --> 00:01:59,520 There are people out there who are very good developers and have lots of programming experience, but 27 00:01:59,520 --> 00:02:01,320 they have no idea what machine learning is. 28 00:02:02,430 --> 00:02:04,740 So all of these people are beginners in some way. 29 00:02:05,250 --> 00:02:09,900 But each of them would require a vastly different amount of work to prepare for this course. 30 00:02:14,940 --> 00:02:17,010 It's the same situation with experts. 31 00:02:17,910 --> 00:02:21,480 Most people consider anyone who knows more than them to be an expert. 32 00:02:22,110 --> 00:02:26,070 But of course, how much you know and how much some other person knows is relative. 33 00:02:26,940 --> 00:02:30,240 If you know how to plug and play into psyche, you learn and nothing else. 34 00:02:30,690 --> 00:02:35,940 You might seem like an expert to someone who only knows excel, but you'll seem like a beginner to someone 35 00:02:35,940 --> 00:02:36,960 who's read bishops. 36 00:02:37,200 --> 00:02:39,240 Pattern recognition and machine learning book. 37 00:02:44,330 --> 00:02:49,010 So just to finalize our answer to this question, is this course for beginners or experts? 38 00:02:49,610 --> 00:02:55,070 This question is misinformed to begin with because beginners and experts are not well defined terms. 39 00:02:55,880 --> 00:03:01,880 Instead, we have to go back to what is the specific skill set you need before starting this course. 40 00:03:02,420 --> 00:03:06,830 And of course, that's defined by the prerequisites which are listed in the course description. 41 00:03:07,370 --> 00:03:09,050 So precision is the key here. 42 00:03:09,470 --> 00:03:15,170 We want to know what is the precise list of skills that I need before taking this course. 43 00:03:15,710 --> 00:03:18,920 This is much better than ambiguous terms like beginner and expert. 44 00:03:23,920 --> 00:03:27,490 Another matter is that people are going to measure their own skills differently. 45 00:03:28,060 --> 00:03:29,920 That's a much tougher issue to deal with. 46 00:03:30,460 --> 00:03:35,770 Sure, I can say you need to know Python coding before starting this course, but you can be sure that 47 00:03:35,770 --> 00:03:41,200 there's a big difference between someone who just learned Helloworld and for loops versus someone working 48 00:03:41,200 --> 00:03:42,790 at a big financial institution. 49 00:03:47,820 --> 00:03:53,130 Now, even in those big financial institutions, there could be big gaps in performance, even within 50 00:03:53,130 --> 00:03:53,860 your own team. 51 00:03:53,880 --> 00:03:55,560 There can be big gaps in performance. 52 00:03:56,160 --> 00:04:01,080 I've seen guys who have been working for 15 years do worse than guys who've been working less than one 53 00:04:01,080 --> 00:04:01,380 year. 54 00:04:02,070 --> 00:04:08,160 So even if I say you need to know python coding, there can be a lot of variation in skill among those 55 00:04:08,160 --> 00:04:09,600 who consider themselves good at it. 56 00:04:10,350 --> 00:04:12,720 A similar situation happens with deep learning. 57 00:04:13,080 --> 00:04:17,670 I can say you need to know how neural networks work, but there's a big difference between someone who 58 00:04:17,670 --> 00:04:22,680 can code them from scratch and has experience using them versus someone who just watched some online 59 00:04:22,680 --> 00:04:24,570 animations for a few hours. 60 00:04:25,230 --> 00:04:28,740 If you had to bet on who knows neural networks better, who would you pick? 61 00:04:29,790 --> 00:04:33,180 So to really answer this question, we have two turns of the course itself. 62 00:04:33,660 --> 00:04:39,180 If I say you need to know Python coding and you claim that you do, but you don't understand some Python 63 00:04:39,180 --> 00:04:44,760 code in this course, then simply put, your Python coding skills are not yet up to par. 64 00:04:45,180 --> 00:04:51,000 So you have to improve your Python coding skills by asking questions on the Q&A and taking the appropriate 65 00:04:51,000 --> 00:04:52,710 steps to cover your gaps. 66 00:04:57,910 --> 00:05:03,400 Quite often, I get people who are just too confident in their own abilities, when unfortunately the 67 00:05:03,400 --> 00:05:09,100 reality is there is a huge discrepancy between what they think they know and what they actually know. 68 00:05:09,940 --> 00:05:13,990 What's funny is these people sometimes say this course requires the IMF. 69 00:05:14,590 --> 00:05:20,980 I assure you, there's absolutely no math in any of my courses that requires a Ph.D. or that only a 70 00:05:21,310 --> 00:05:22,390 Ph.D. can understand. 71 00:05:22,870 --> 00:05:26,080 My courses contain only undergraduate math at most. 72 00:05:26,860 --> 00:05:32,140 Sometimes people even claim to have a Ph.D. and then claim that it's still too hard, which is funny 73 00:05:32,140 --> 00:05:35,060 because that essentially invalidates your entire Ph.D. 74 00:05:35,890 --> 00:05:40,060 The entire point of a Ph.D. is to turn you into an independent researcher. 75 00:05:40,390 --> 00:05:45,010 And so if you're supposed to be at the level of doing independent research, but you can't understand 76 00:05:45,010 --> 00:05:47,800 some undergraduate math, well, that's not very good. 77 00:05:52,960 --> 00:05:58,510 Essentially, what it boils down to is people are really trying to say everything I don't understand 78 00:05:58,510 --> 00:05:59,320 is not important. 79 00:05:59,800 --> 00:06:05,500 Everything I do understand is important and therefore everything you say that I don't understand, I 80 00:06:05,500 --> 00:06:06,220 don't need to know. 81 00:06:06,910 --> 00:06:08,980 You can see why this thinking is dangerous. 82 00:06:09,490 --> 00:06:14,440 If you truly believe that everything you do not know is not important, then you'll never learn anything 83 00:06:14,440 --> 00:06:14,800 new. 84 00:06:19,830 --> 00:06:24,300 Now this brings us to our third question, is this course fast paced or slow paced? 85 00:06:24,930 --> 00:06:26,850 And I think you can see where I'm going with this. 86 00:06:27,360 --> 00:06:33,990 This is yet another example of ambiguity fast pace to whom, whether it's fast pace or not, depends 87 00:06:33,990 --> 00:06:35,880 on which person is taking the course. 88 00:06:36,450 --> 00:06:39,780 So if someone is well prepared, then things will be just right. 89 00:06:40,500 --> 00:06:44,520 If someone already knows most of these topics, then it will probably be too slow. 90 00:06:45,540 --> 00:06:50,910 If someone chose not to prepare for the course or they have an inadequate understanding of the prerequisites, 91 00:06:51,240 --> 00:06:52,800 then it's going to be too fast. 92 00:06:53,490 --> 00:06:58,080 So if I expected you to know something and you didn't, then you have to look it up, which is going 93 00:06:58,080 --> 00:07:00,240 to make the course feel like it's going too fast. 94 00:07:05,320 --> 00:07:06,760 So here is the real test. 95 00:07:07,360 --> 00:07:13,120 If I say something depends on calculus of probability and you claim to know calculus and probability, 96 00:07:13,510 --> 00:07:18,460 but you don't understand the thing that depends on calculus and probability, you have to ask yourself, 97 00:07:18,460 --> 00:07:21,970 honestly, do I really understand calculus and probability? 98 00:07:22,450 --> 00:07:27,700 Or did the instructor invent his own kind of calculus and probability that I don't yet understand? 99 00:07:28,180 --> 00:07:29,920 Well, of course, that's not very likely. 100 00:07:30,430 --> 00:07:35,320 It's most likely that you thought you understood calculus and probability, but you really don't. 101 00:07:40,350 --> 00:07:43,950 Now onto our initial question, what kind, of course, is this? 102 00:07:44,040 --> 00:07:45,060 Is it academic? 103 00:07:45,390 --> 00:07:46,610 Is it for professionals? 104 00:07:46,620 --> 00:07:47,520 Is it practical? 105 00:07:48,000 --> 00:07:53,310 So to understand this, you have to understand where machine learning is usually taught and how it is 106 00:07:53,310 --> 00:07:53,940 usually taught. 107 00:07:55,010 --> 00:08:00,050 Firstly, machine learning is usually taught in upper year computer science and engineering programs 108 00:08:00,260 --> 00:08:01,820 in university and college. 109 00:08:02,390 --> 00:08:07,340 So these are students who already took calculus one to three linear algebra, differential equations, 110 00:08:07,550 --> 00:08:13,700 probability and statistics, discrete math programming and possibly more all in their first two or three 111 00:08:13,700 --> 00:08:14,570 years of college. 112 00:08:15,170 --> 00:08:19,460 So by the time they get to third and fourth year, they know all this stuff and they're ready to apply 113 00:08:19,460 --> 00:08:19,580 it. 114 00:08:24,720 --> 00:08:30,390 So a typical academic course might cover a whole library of machine learning algorithms, so you might 115 00:08:30,390 --> 00:08:36,809 learn closest neighbor logistic regression K means clustering PCA and so on, all in the same course. 116 00:08:37,840 --> 00:08:42,559 Typically, you'd cover one algorithm per lecture, which is about two hours overall. 117 00:08:42,580 --> 00:08:48,190 You might have something like 12 or maybe 20 of those lectures in a term and by cover, I mean, mostly 118 00:08:48,190 --> 00:08:50,890 via math, geometry and derivations. 119 00:08:51,430 --> 00:08:55,090 These courses don't typically have you at your computer doing programming work. 120 00:08:55,900 --> 00:09:00,820 If you do end up doing programming, it's usually part of a lab assignment, which is actually just 121 00:09:00,820 --> 00:09:03,820 a very small part of the course compared to the rest. 122 00:09:04,510 --> 00:09:10,000 So it's mostly theoretical dealing with equations, solving equations, reasoning about equations. 123 00:09:10,450 --> 00:09:14,200 Those equations typically involve probabilities and information theory. 124 00:09:14,710 --> 00:09:18,130 So you have two hours to go through the theory for one algorithm. 125 00:09:18,130 --> 00:09:22,870 And then if there's coding involved, which there usually isn't, you do it on your own time. 126 00:09:23,560 --> 00:09:26,110 The lectures themselves do not involve any coding. 127 00:09:31,160 --> 00:09:34,100 But there's an even bigger problem with these academic courses. 128 00:09:34,240 --> 00:09:34,940 And what is that? 129 00:09:35,870 --> 00:09:38,510 What happens is that the grades turn out to be very low. 130 00:09:41,060 --> 00:09:45,620 Out of 100 percent, the average might be around 25 percent or up to 40 percent. 131 00:09:46,220 --> 00:09:48,440 Now you might think, wow, so everyone fails. 132 00:09:48,890 --> 00:09:53,270 And of course, that's not the case since everyone gets shifted based on the average. 133 00:09:53,900 --> 00:09:55,490 So what is the real problem here? 134 00:09:56,150 --> 00:10:01,040 A lot of people ask me, Well, why don't you cover every machine learning algorithm in one big mass, 135 00:10:01,040 --> 00:10:03,230 of course, and just call it machine learning? 136 00:10:04,010 --> 00:10:09,860 Well, the answer is that if you tried to do too much, you only end up understanding 25 percent of 137 00:10:09,860 --> 00:10:12,650 the material, which I think is clearly demonstrated here. 138 00:10:13,520 --> 00:10:15,500 And as they say, the proof is in the pudding. 139 00:10:16,010 --> 00:10:20,780 If you look at these courses where students are trying to digest 10 different algorithms at the same 140 00:10:20,780 --> 00:10:22,400 time, what is the result? 141 00:10:22,940 --> 00:10:24,980 The result is they don't understand any of them. 142 00:10:25,400 --> 00:10:29,120 If you try to understand too many things, you end up understanding nothing. 143 00:10:30,050 --> 00:10:35,750 I take the perspective that if you're going to take a course, you should aim to understand 100 percent 144 00:10:35,750 --> 00:10:36,620 of the material. 145 00:10:37,880 --> 00:10:43,460 Each course is broken up into appropriate levels based on how difficult they are and by topic. 146 00:10:43,940 --> 00:10:50,030 So for example, you can typically take Calculus one, which focuses on differential calculus calculus 147 00:10:50,030 --> 00:10:55,880 two, which focuses on integral calculus and calculus three, which focuses on multivariable calculus. 148 00:10:56,420 --> 00:11:01,340 You would never learn all of these in just one huge course called math, because that's just too much 149 00:11:01,340 --> 00:11:03,410 stuff to understand all at once. 150 00:11:04,040 --> 00:11:10,790 I want you to understand 100 percent of the material, even 50 percent or 75 percent is not good enough, 151 00:11:10,790 --> 00:11:15,140 in my opinion, if you only understand 75 percent of the material. 152 00:11:15,380 --> 00:11:19,460 And the next course depends on the 25 percent that you didn't understand. 153 00:11:19,760 --> 00:11:22,160 Well, then you're not going to make it to the next level. 154 00:11:27,200 --> 00:11:29,630 Now, let's look at the opposite end of the spectrum. 155 00:11:30,020 --> 00:11:32,780 What most people might consider to be a practical course. 156 00:11:33,780 --> 00:11:37,110 I don't consider these to be practical courses, and I'll explain why later. 157 00:11:37,470 --> 00:11:42,270 But I use the word practical here in quotes, since that's what some of the more beginner students will 158 00:11:42,270 --> 00:11:43,170 refer to the mess. 159 00:11:43,980 --> 00:11:46,380 So how does a practical course typically proceed? 160 00:11:47,970 --> 00:11:52,950 Well, first, you are never really taught how the algorithms work other than at a very high level. 161 00:11:53,520 --> 00:11:56,880 You might learn how they work, but we have analogies and metaphors. 162 00:11:57,360 --> 00:12:00,600 So, for example, suppose you're learning about gradient descent. 163 00:12:01,170 --> 00:12:04,560 Well, the instructor might relate this to a ball rolling down a hill. 164 00:12:05,850 --> 00:12:10,680 Now, some beginner students, they might conclude that I understand a ball rolling down a hill. 165 00:12:10,860 --> 00:12:16,020 Therefore, I understand gradient descent, but they neglect to understand how to take derivatives. 166 00:12:16,260 --> 00:12:18,630 They neglect to understand how to choose a good learning rate. 167 00:12:19,020 --> 00:12:22,620 And most importantly, they neglect how to implement gradient descent in code. 168 00:12:24,160 --> 00:12:29,200 So it's totally wrong to assume that just because you can visualize a ball rolling down a hill that 169 00:12:29,200 --> 00:12:31,690 you understand gradient descent, why? 170 00:12:32,840 --> 00:12:38,060 Well, the problem is, everybody can imagine a ball rolling down a hill, your grandparents who can't 171 00:12:38,060 --> 00:12:41,600 even turn on a computer, can't understand a ball rolling down a hill. 172 00:12:41,990 --> 00:12:46,550 So don't equate understanding an analogy to understanding the real topic. 173 00:12:51,790 --> 00:12:57,550 So what's the next thing these practical courses do, the next thing they usually teach you is how to 174 00:12:57,550 --> 00:13:02,380 plug into an API or, in other words, make use of a library, someone else wrote. 175 00:13:02,980 --> 00:13:04,570 This is also not ideal. 176 00:13:05,200 --> 00:13:07,810 Typically, this involves just two or three lines of code. 177 00:13:08,410 --> 00:13:13,060 You have to wonder if machine learning just boils down to two or three lines of code. 178 00:13:13,420 --> 00:13:15,190 Why is it such a big deal these days? 179 00:13:15,580 --> 00:13:18,610 Why did it take decades for people to realize its usefulness? 180 00:13:19,210 --> 00:13:23,470 And of course, the answer is that there's much more to machine learning than just these two or three 181 00:13:23,470 --> 00:13:24,190 lines of code. 182 00:13:24,880 --> 00:13:29,110 When you use a deep learning library like Keros, you're not doing deep learning. 183 00:13:29,470 --> 00:13:31,600 You were using a deep learning interface. 184 00:13:31,750 --> 00:13:33,010 You're simply programming. 185 00:13:33,430 --> 00:13:39,580 It's more accurate to say I can program with deep learning libraries rather than saying I actually know 186 00:13:39,580 --> 00:13:40,180 deep learning. 187 00:13:44,860 --> 00:13:47,500 But some of you might say, well, that's all I really want to do. 188 00:13:47,710 --> 00:13:51,100 All I need is an API and for the API to give me an answer. 189 00:13:51,520 --> 00:13:53,050 I don't care about how it works. 190 00:13:53,560 --> 00:13:55,450 But then here is where you run into trouble. 191 00:13:56,140 --> 00:13:58,750 Recall that in addition to all data is the same. 192 00:13:59,080 --> 00:14:01,930 We know that all machine learning interfaces are the same. 193 00:14:02,470 --> 00:14:06,580 So the question becomes Which API do you choose if they're all the same? 194 00:14:07,330 --> 00:14:10,420 How can you make an informed choice about which one to use? 195 00:14:11,080 --> 00:14:15,910 Well, the reality is, if you don't know how these things work, then you can't answer this question. 196 00:14:16,540 --> 00:14:21,820 This is why some of these so-called practical courses can cover 20 algorithms in the same course. 197 00:14:22,180 --> 00:14:25,390 It's because they never actually talk about how the algorithms work. 198 00:14:26,080 --> 00:14:28,120 Since the API is the same every time. 199 00:14:28,300 --> 00:14:32,560 It's very easy to just do the same thing 20 times and then not bother with the details. 200 00:14:33,310 --> 00:14:37,510 So this gives people a lot of confidence that they were able to do the same thing 20 times. 201 00:14:38,050 --> 00:14:42,740 But it's dangerous thinking because you don't realize you really only learned one thing. 202 00:14:42,760 --> 00:14:44,470 You just repeated it 20 times. 203 00:14:49,530 --> 00:14:51,990 So how about when it comes to the professional world? 204 00:14:52,930 --> 00:14:58,390 Sometimes people tend to believe that in your profession, you'll be leaning more toward the practical 205 00:14:58,390 --> 00:15:02,800 end of the spectrum, where you don't really need to know how a machine learning model works. 206 00:15:03,040 --> 00:15:04,570 You just have to know its API. 207 00:15:05,500 --> 00:15:08,710 But what it really comes down to is what your job actually is. 208 00:15:09,220 --> 00:15:13,900 If you're a regular everyday programmer and you just want to see how a machine learning model might 209 00:15:13,900 --> 00:15:17,350 perform on your data, then you're going to use a pre-built library. 210 00:15:17,350 --> 00:15:18,070 Most likely. 211 00:15:18,490 --> 00:15:21,250 That's just the two or three lines of code I referenced earlier. 212 00:15:22,780 --> 00:15:28,180 However, that does not mean you don't still have to conform to best practices, make sure you're not 213 00:15:28,180 --> 00:15:30,040 mixing your test set with your train set. 214 00:15:30,460 --> 00:15:33,040 Make sure you're doing proper validation and so on. 215 00:15:33,550 --> 00:15:38,260 And when it comes to tuning your model, then you'll have to be able to dig down into the details and 216 00:15:38,260 --> 00:15:42,790 see which hyper parameters are possible to change and which are likely to have the most impact. 217 00:15:43,780 --> 00:15:48,250 So only knowing how to use an API is only going to get you past the first stage. 218 00:15:48,640 --> 00:15:52,600 You can plug in your data and at least your code is the right syntax and doesn't crash. 219 00:15:52,960 --> 00:15:56,710 But of course, your code not crashing doesn't necessarily mean it's right. 220 00:16:01,680 --> 00:16:06,630 Back on the other end of the spectrum, you have programmers and scientists whose sole job it is to 221 00:16:06,630 --> 00:16:07,530 do machine learning. 222 00:16:08,040 --> 00:16:09,930 These guys are machine learning experts. 223 00:16:10,350 --> 00:16:14,070 They know the math inside and out and have a full computer science education. 224 00:16:15,000 --> 00:16:20,370 And so to even be able to communicate your ideas effectively, you have to be comfortable with the theory 225 00:16:20,370 --> 00:16:23,400 and not just how to plug into an API with two lines of code. 226 00:16:24,300 --> 00:16:29,880 These scientists are building custom models, really digging down into the math to not only use but 227 00:16:29,880 --> 00:16:33,360 invent new things and to be able to invent new things. 228 00:16:33,600 --> 00:16:35,610 You really need to know the math very well. 229 00:16:36,240 --> 00:16:40,260 And so if you're on a team like this, you can't just know how to write two lines of code. 230 00:16:40,560 --> 00:16:43,020 Your skill set has to be far beyond that. 231 00:16:43,980 --> 00:16:49,260 So your theoretical background has to be very solid if you want to be able to communicate effectively 232 00:16:49,410 --> 00:16:50,580 with other scientists. 233 00:16:55,770 --> 00:16:59,880 Now, a lot of people say, well, I don't ever want to be a scientist, so I don't really care how 234 00:16:59,880 --> 00:17:00,900 these algorithms work. 235 00:17:01,560 --> 00:17:05,220 Unfortunately, that's a very bad attitude to have in the workplace. 236 00:17:05,849 --> 00:17:10,150 Suppose, for example, I am running a business now, an employee of mine. 237 00:17:10,170 --> 00:17:13,140 He wants to show off and say he's learning about machine learning. 238 00:17:13,650 --> 00:17:16,319 He says, Oh, I'm taking this machine learning course online. 239 00:17:16,589 --> 00:17:18,960 Therefore, I'll become a machine learning master. 240 00:17:19,619 --> 00:17:23,040 But his approach to machine learning is to learn as little as possible. 241 00:17:23,430 --> 00:17:25,050 He says he doesn't care about the math. 242 00:17:25,079 --> 00:17:26,579 He only wants the practical stuff. 243 00:17:26,940 --> 00:17:28,980 He only wants to know the bare minimum. 244 00:17:29,640 --> 00:17:31,550 He tells me he doesn't care how it works. 245 00:17:31,560 --> 00:17:36,200 He's just going to plug into some API because that's what his buddy told him he could do. 246 00:17:37,270 --> 00:17:40,810 As a business owner, I really don't want this guy working for me. 247 00:17:41,440 --> 00:17:47,140 I run a business that sells something and the life of the company itself and the livelihoods of the 248 00:17:47,140 --> 00:17:51,430 employees all depend on our products being the best they can possibly be. 249 00:17:51,970 --> 00:17:55,210 So if someone comes to me and says I don't care about being the best. 250 00:17:55,510 --> 00:17:57,280 I just care about doing the bare minimum. 251 00:17:57,730 --> 00:18:00,160 My response is I don't want that guy on my team. 252 00:18:00,490 --> 00:18:01,900 In fact, he should be fired. 253 00:18:02,560 --> 00:18:05,080 So be careful if this is your approach to machine learning. 254 00:18:05,620 --> 00:18:10,720 No business owner wants to hear that your approach to the business is to do the bare minimum and to 255 00:18:10,720 --> 00:18:12,940 put off the math because you think it's too hard. 256 00:18:13,450 --> 00:18:15,590 That's a scary prospect for a business owner. 257 00:18:20,640 --> 00:18:25,020 Now, if we go back to our original question, maybe you've realized we have an answer to yet. 258 00:18:25,650 --> 00:18:27,060 What kind, of course, is this? 259 00:18:27,720 --> 00:18:30,780 Well, you can think of it as a marriage between these approaches. 260 00:18:31,470 --> 00:18:35,310 We want to cover all the math and derivations because that's very important. 261 00:18:35,670 --> 00:18:40,530 If you want to be able to communicate effectively with other scientists to be able to debug your code 262 00:18:40,530 --> 00:18:46,590 effectively and to be able to invent new approaches, this is also what we mean by machine learning. 263 00:18:47,100 --> 00:18:51,180 We mean really knowing machine learning and not just programming with an API. 264 00:18:51,990 --> 00:18:54,330 So the theoretical background is very important. 265 00:18:54,810 --> 00:19:00,630 At the same time, we're not going to introduce 20 different algorithms in the same course because unlike 266 00:19:00,630 --> 00:19:06,840 a typical academic course where you can pass just by understanding 25 percent of the material in this 267 00:19:06,840 --> 00:19:12,120 course, it's structured in a way that if you have any missing gaps, it's going to be impossible for 268 00:19:12,120 --> 00:19:13,680 you to move on to the next stage. 269 00:19:15,080 --> 00:19:18,140 But academic courses don't contain much coding, if at all. 270 00:19:18,620 --> 00:19:22,880 So in this course, every algorithm we learn about will be implemented in code. 271 00:19:23,480 --> 00:19:27,080 My motto is if you can't implement it, then you don't understand it. 272 00:19:27,830 --> 00:19:29,360 You can pretend to understand it. 273 00:19:29,630 --> 00:19:32,240 And when you talk about it, you might sound like you understand it. 274 00:19:32,510 --> 00:19:34,580 But implementation is the true test. 275 00:19:35,060 --> 00:19:36,170 It's like a math exam. 276 00:19:36,170 --> 00:19:41,360 Rather than writing an essay when you write an essay, you might be able to fool the teacher into thinking 277 00:19:41,360 --> 00:19:42,440 you understand something. 278 00:19:42,830 --> 00:19:46,790 But in a math exam, either your answers are right or your answers are wrong. 279 00:19:47,330 --> 00:19:50,000 In this context, either your code works or it doesn't. 280 00:19:50,180 --> 00:19:51,230 It's a very clear cut. 281 00:19:55,910 --> 00:19:58,490 When it comes to practicality, we cover that, too. 282 00:19:59,030 --> 00:20:03,350 We demonstrate our algorithms both on real world data like text and images. 283 00:20:03,620 --> 00:20:08,930 And we demonstrate our algorithms on two dimensional data sets so that you can visualize what an algorithm 284 00:20:08,930 --> 00:20:09,410 is doing. 285 00:20:10,190 --> 00:20:11,420 Both of these are important. 286 00:20:12,290 --> 00:20:16,040 Texts and images are quite possibly the most practical data there is. 287 00:20:16,460 --> 00:20:20,660 Some billion dollar companies exist and thrive because of those capabilities. 288 00:20:21,470 --> 00:20:26,180 2D data sets are critical as well because as humans, that's the only thing we can see. 289 00:20:26,720 --> 00:20:31,460 But at the same time, we don't try to pull the wool over your eyes and pretend like you need 20 different 290 00:20:31,460 --> 00:20:32,750 data sets to practice on. 291 00:20:33,290 --> 00:20:35,460 Instead, we learn something much more valuable. 292 00:20:35,810 --> 00:20:41,000 How to use an algorithm on an infinite number of data sets because we are able to abstract the idea 293 00:20:41,000 --> 00:20:42,620 that all data is the same. 294 00:20:43,280 --> 00:20:45,910 Remember that all the computer sees is a list of numbers. 295 00:20:45,920 --> 00:20:51,290 It doesn't know that those numbers represent height measurements or an image pixel, and of course, 296 00:20:51,290 --> 00:20:56,060 by experiencing both ends of the spectrum, both the theory and the implementation. 297 00:20:56,330 --> 00:21:01,340 And then finally, plugging data into your model that's going to make you very well-rounded professionally. 298 00:21:06,580 --> 00:21:11,950 To conclude this lecture, I want to pose a final question, how does all this knowledge help you? 299 00:21:12,610 --> 00:21:17,170 Why does understanding the approach of this course make you better at machine learning in general? 300 00:21:18,070 --> 00:21:22,360 Well, my hope is that this gives you a very high level bird's eye view of the landscape. 301 00:21:22,900 --> 00:21:26,140 You can see the different approaches of people from different backgrounds. 302 00:21:26,920 --> 00:21:32,050 Academic people typically don't see what other non-academic people are doing, and those people who 303 00:21:32,050 --> 00:21:35,500 hate math typically don't see what people who love math are doing. 304 00:21:35,950 --> 00:21:38,440 So with this bird's eye view, you can see everybody. 305 00:21:39,130 --> 00:21:43,840 If you're learning machine learning, it helps you understand who am I competing with when I apply for 306 00:21:43,840 --> 00:21:44,380 a job? 307 00:21:44,770 --> 00:21:46,300 And what kinds of jobs are out there? 308 00:21:46,300 --> 00:21:48,190 And what would be adequate preparation? 309 00:21:48,760 --> 00:21:53,290 If you're hiring for a machine learning team, it helps you understand the different backgrounds of 310 00:21:53,290 --> 00:21:54,070 individuals. 311 00:21:54,070 --> 00:21:58,720 You have to compare to one another to understand the strengths and weaknesses and what they bring to 312 00:21:58,720 --> 00:21:59,200 the table.