1 00:00:00,000 --> 00:00:04,572 [MUSIC] 2 00:00:04,572 --> 00:00:09,020 This course is going to follow the same philosophy as our past courses. 3 00:00:09,020 --> 00:00:12,770 In particular, we're going to use case studies to motivate the key concepts that 4 00:00:12,770 --> 00:00:14,350 we're going to teach. 5 00:00:14,350 --> 00:00:17,850 But there are a number of other key features that define the way we teach our 6 00:00:17,850 --> 00:00:20,090 courses in this specialization. 7 00:00:20,090 --> 00:00:24,230 In particular, we teach a set of core machine learning concepts and 8 00:00:24,230 --> 00:00:26,940 we do so both through our case studies 9 00:00:26,940 --> 00:00:31,300 as well as using visual aids to very dramatically guide the process. 10 00:00:32,340 --> 00:00:37,430 And then, in these courses following on from the foundations course, we're 11 00:00:37,430 --> 00:00:42,700 going to go into details on the algorithms of the methods provided in the course. 12 00:00:42,700 --> 00:00:46,290 And we're not just going to provide a laundry list of methods people use for 13 00:00:46,290 --> 00:00:47,320 clustering and retrieval. 14 00:00:47,320 --> 00:00:51,760 We're going to focus in on the methods we feel that are most widely used. 15 00:00:51,760 --> 00:00:53,830 The most practical algorithms out there. 16 00:00:53,830 --> 00:00:56,550 And the ones that give us the most skills for 17 00:00:56,550 --> 00:01:01,240 learning other algorithms that might exist now or in the future. 18 00:01:01,240 --> 00:01:02,330 And throughout this course, 19 00:01:02,330 --> 00:01:06,600 you're going to get hands on experience with implementing these methods. 20 00:01:06,600 --> 00:01:08,780 And not only are you going to implement these methods, but 21 00:01:08,780 --> 00:01:12,170 you're going to do so on these real world applications. 22 00:01:12,170 --> 00:01:15,320 So, you're going to get actual experience of deploying these 23 00:01:15,320 --> 00:01:17,900 machine learning algorithms on data sets 24 00:01:17,900 --> 00:01:21,370 that are things one might actually consider out there in the world. 25 00:01:22,440 --> 00:01:25,750 And through this process you're going to just gain a lot of intuition 26 00:01:25,750 --> 00:01:29,630 about the methods and there potential limitations and strengths. 27 00:01:30,860 --> 00:01:35,370 And finally we also teach a set of advanced concepts in this course 28 00:01:35,370 --> 00:01:36,640 that we mark as optional. 29 00:01:36,640 --> 00:01:40,340 So these are videos that if you're interested in watching some of the details 30 00:01:40,340 --> 00:01:43,980 that are under the hood and some of the things that we're going to describe, 31 00:01:43,980 --> 00:01:47,130 you can watch these videos, but if you prefer not to that's totally fine. 32 00:01:47,130 --> 00:01:51,180 You're going to get a very, very thorough overview of clustering and retrieval. 33 00:01:51,180 --> 00:01:55,280 You'll still be able to implement and deploy these methods but 34 00:01:55,280 --> 00:02:00,330 you might just not understand some of the proofs or the more detailed concepts but 35 00:02:00,330 --> 00:02:02,890 this content is here for those that are interested in it. 36 00:02:04,180 --> 00:02:06,260 Well, more specifically in this course, 37 00:02:06,260 --> 00:02:09,530 we're going to go through a number of different models. 38 00:02:09,530 --> 00:02:12,140 Like nearest neighbors for search. 39 00:02:12,140 --> 00:02:15,960 We're going to talk about clustering as a high-level task, 40 00:02:15,960 --> 00:02:17,840 an unsupervised learning task. 41 00:02:17,840 --> 00:02:21,360 And we're going to talk about problemistic models for performing clusters. 42 00:02:21,360 --> 00:02:24,100 Clustering like mixture models. 43 00:02:24,100 --> 00:02:27,120 And then we're going to talk about a more intricate problemistic model called 44 00:02:27,120 --> 00:02:28,430 latent Dirichlet allocation. 45 00:02:29,680 --> 00:02:32,580 Then we are going to go through a number of algorithms associated with 46 00:02:32,580 --> 00:02:33,315 these models. 47 00:02:33,315 --> 00:02:37,760 KD-trees as an efficient data structure for 48 00:02:37,760 --> 00:02:40,825 performing our nearest neighbor search, locality sensitive hashing, 49 00:02:40,825 --> 00:02:45,310 k-means as a way of doing our clustering as well as MapReduce which we 50 00:02:45,310 --> 00:02:49,820 mentioned is a means of paralyzing our algorithms to scale them up. 51 00:02:49,820 --> 00:02:53,530 Expectation maximization for inference in our mixture models and 52 00:02:53,530 --> 00:02:58,040 finally, Gibbs sampling for inference in our Latent Dirichlet allocation model. 53 00:02:59,270 --> 00:03:02,550 And importantly, throughout the course, we're going to cover a number 54 00:03:02,550 --> 00:03:07,600 of fundamental machine learning concepts that extent beyond just the concepts of 55 00:03:07,600 --> 00:03:11,410 clustering and retrieval We're going to talk about distance metrics. 56 00:03:11,410 --> 00:03:15,240 We're going to talk about approximation, algorithms, unsupervised learning, 57 00:03:16,700 --> 00:03:21,550 probabilistic modeling, data parallel problems, and Bayesian inference. 58 00:03:21,550 --> 00:03:24,370 So really a wide range of concepts covered in this course.