1 00:00:00,000 --> 00:00:02,395 Hello and welcome. 2 00:00:02,395 --> 00:00:03,810 As you probably know, 3 00:00:03,810 --> 00:00:05,640 deep learning has already transformed 4 00:00:05,640 --> 00:00:09,325 traditional internet businesses like web search and advertising. 5 00:00:09,325 --> 00:00:11,160 But deep learning is also enabling 6 00:00:11,160 --> 00:00:15,085 brand new products and businesses and ways of helping people to be created. 7 00:00:15,085 --> 00:00:17,205 Everything ranging from better healthcare, 8 00:00:17,205 --> 00:00:19,410 where deep learning is getting really good at reading 9 00:00:19,410 --> 00:00:22,935 X-ray images to delivering personalized education, 10 00:00:22,935 --> 00:00:28,320 to precision agriculture, to even self driving cars and many others. 11 00:00:28,320 --> 00:00:30,790 If you want to learn the tools of deep learning and 12 00:00:30,790 --> 00:00:33,495 be able to apply them to build these amazing things, 13 00:00:33,495 --> 00:00:35,675 I want to help you get there. 14 00:00:35,675 --> 00:00:38,725 When you finish the sequence of courses on Coursera, 15 00:00:38,725 --> 00:00:41,190 called the specialization, you will be able to put 16 00:00:41,190 --> 00:00:44,455 deep learning onto your resume` with confidence. 17 00:00:44,455 --> 00:00:45,615 Over the next decade, 18 00:00:45,615 --> 00:00:50,430 I think all of us have an opportunity to build an amazing world, amazing society, 19 00:00:50,430 --> 00:00:52,004 that is AI powers, 20 00:00:52,004 --> 00:00:57,364 and I hope that you will play a big role in the creation of this AI power society. 21 00:00:57,364 --> 00:01:00,782 So that, let's get started. 22 00:01:00,782 --> 00:01:03,690 I think that AI is the new electricity. 23 00:01:03,690 --> 00:01:05,485 Starting about 100 years ago, 24 00:01:05,485 --> 00:01:10,055 the electrification of our society transformed every major industry, 25 00:01:10,055 --> 00:01:12,430 every ranging from transportation, manufacturing, 26 00:01:12,430 --> 00:01:15,430 to healthcare, to communications and many more. 27 00:01:15,430 --> 00:01:18,280 And today, we see a surprisingly clear path 28 00:01:18,280 --> 00:01:22,225 for AI to bring about an equally big transformation. 29 00:01:22,225 --> 00:01:25,090 And of course, the part of AI that is rising 30 00:01:25,090 --> 00:01:29,530 rapidly and driving a lot of these developments, is deep learning. 31 00:01:29,530 --> 00:01:31,630 So today, deep learning is one of the most 32 00:01:31,630 --> 00:01:34,920 highly sought after skills and technology worlds. 33 00:01:34,920 --> 00:01:37,795 And through this course and a few causes after this one, 34 00:01:37,795 --> 00:01:40,900 I want to help you to gain and master those skills. 35 00:01:40,900 --> 00:01:43,870 So here's what you learn in this sequence of 36 00:01:43,870 --> 00:01:47,290 courses also called a specialization on Coursera. 37 00:01:47,290 --> 00:01:48,775 In the first course, 38 00:01:48,775 --> 00:01:51,925 you learn about the foundations of neural networks, 39 00:01:51,925 --> 00:01:54,555 you learn about neural networks and deep learning. 40 00:01:54,555 --> 00:01:56,650 This video that you're watching is part of 41 00:01:56,650 --> 00:01:59,895 this first course which last four weeks in total. 42 00:01:59,895 --> 00:02:05,450 And each of the five courses in the specialization will be about two to four weeks, 43 00:02:05,450 --> 00:02:08,490 with most of them actually shorter than four weeks. 44 00:02:08,490 --> 00:02:09,685 But in this first course, 45 00:02:09,685 --> 00:02:12,340 you'll learn how to build a new network including 46 00:02:12,340 --> 00:02:15,520 a deep neural network and how to train it on data. 47 00:02:15,520 --> 00:02:16,840 And at the end of this course, 48 00:02:16,840 --> 00:02:21,445 you'll be able to build a deep neural network to recognize, guess what? 49 00:02:21,445 --> 00:02:23,710 Cats. For some reason, 50 00:02:23,710 --> 00:02:27,370 there is a cat neem running around in deep learning. 51 00:02:27,370 --> 00:02:30,020 And so, following tradition in this first course, 52 00:02:30,020 --> 00:02:32,700 we'll build a cat recognizer. 53 00:02:32,700 --> 00:02:35,395 Then in the second course, 54 00:02:35,395 --> 00:02:38,420 you learn about the practical aspects of deep learning. 55 00:02:38,420 --> 00:02:40,810 So you learn, now that you've built in your network, 56 00:02:40,810 --> 00:02:42,765 how to actually get it to perform well. 57 00:02:42,765 --> 00:02:46,295 So you learn about hyperparameter tuning, regularization, 58 00:02:46,295 --> 00:02:49,060 how to diagnose price and variants and advance 59 00:02:49,060 --> 00:02:54,755 optimization algorithms like momentum armrest prop and the ad authorization algorithm. 60 00:02:54,755 --> 00:02:56,645 Sometimes it seems like there's a lot of tuning, 61 00:02:56,645 --> 00:02:59,925 even some black magic and how you build a new network. 62 00:02:59,925 --> 00:03:02,740 So the second course which is just three weeks, 63 00:03:02,740 --> 00:03:04,900 will demystify some of that black magic. 64 00:03:04,900 --> 00:03:07,710 In the third course which is just two weeks, 65 00:03:07,710 --> 00:03:10,930 you learn how to structure your machine learning project. 66 00:03:10,930 --> 00:03:13,270 It turns out that the strategy for building 67 00:03:13,270 --> 00:03:16,820 a machine learning system has changed in the era of deep learning. 68 00:03:16,820 --> 00:03:20,785 So for example, the way you switch your data into train, 69 00:03:20,785 --> 00:03:26,095 development or dev also called holdout cross-validation sets and test sets, 70 00:03:26,095 --> 00:03:29,005 has changed in the era of deep learning. 71 00:03:29,005 --> 00:03:31,860 So whether the new best practices are doing that 72 00:03:31,860 --> 00:03:36,215 and whether if you were training set and your test come from different distributions, 73 00:03:36,215 --> 00:03:38,760 that's happening a lot more in the era of deep learning. 74 00:03:38,760 --> 00:03:40,710 So how do you deal with that? 75 00:03:40,710 --> 00:03:44,700 And if you've heard of end to end deep learning, 76 00:03:44,700 --> 00:03:48,375 you also learn more about that in this third course 77 00:03:48,375 --> 00:03:51,690 and see when you should use it and maybe when you shouldn't. 78 00:03:51,690 --> 00:03:54,960 The material in this third course is relatively unique. 79 00:03:54,960 --> 00:03:57,990 I'm going to share of you a lot of the hard one lessons that I've learned, 80 00:03:57,990 --> 00:04:01,860 building and shipping, quite a lot of deep learning products. 81 00:04:01,860 --> 00:04:03,075 As far as I know, 82 00:04:03,075 --> 00:04:06,420 this is largely material that is not taught in 83 00:04:06,420 --> 00:04:10,650 most universities that have deep learning courses. 84 00:04:10,650 --> 00:04:15,266 But I really hope you to get your deep learning systems to work well. 85 00:04:15,266 --> 00:04:16,755 In the next course, 86 00:04:16,755 --> 00:04:22,530 we'll then talk about convolutional neural networks, often abbreviated CNNs. 87 00:04:22,530 --> 00:04:28,670 Convolutional networks or convolutional neural networks are often applied to images. 88 00:04:28,670 --> 00:04:31,440 So you learn how to build these models in course four. 89 00:04:31,440 --> 00:04:33,435 Finally, in course five, 90 00:04:33,435 --> 00:04:36,300 you learn sequence models and how to apply 91 00:04:36,300 --> 00:04:40,080 them to natural language processing and other problems. 92 00:04:40,080 --> 00:04:42,600 So sequence models includes models like 93 00:04:42,600 --> 00:04:47,465 recurrent neural networks abbreviated RNNs and LSTM models, 94 00:04:47,465 --> 00:04:49,645 sense for a long short term memory models. 95 00:04:49,645 --> 00:04:52,740 You'll learn what these terms mean in course five and be able to 96 00:04:52,740 --> 00:04:56,865 apply them to natural language processing problems. 97 00:04:56,865 --> 00:05:02,220 So you learn these models in course five and be able to apply them to sequence data. 98 00:05:02,220 --> 00:05:05,670 So for example, natural language is just a sequence of words, 99 00:05:05,670 --> 00:05:10,505 and you also understand how these models can be applied to speech recognition, 100 00:05:10,505 --> 00:05:14,140 or to music generation, and other problems. 101 00:05:14,140 --> 00:05:15,735 So through these courses, 102 00:05:15,735 --> 00:05:17,705 you'll learn the tools of deep learning, 103 00:05:17,705 --> 00:05:20,100 you'll be able to apply them to build amazing things, 104 00:05:20,100 --> 00:05:24,810 and I hope many of you through this will also be able to advance your career. 105 00:05:24,810 --> 00:05:26,910 So that, let's get started. 106 00:05:26,910 --> 00:05:29,145 Please go on to the next video where we'll talk about 107 00:05:29,145 --> 00:05:31,650 deep learning applied to supervise learning.