1 00:00:00,210 --> 00:00:00,780 Hello. 2 00:00:00,930 --> 00:00:06,630 So before going deep down into the session, let's have a quick recap of what we all have learned in 3 00:00:06,630 --> 00:00:07,760 our previous session. 4 00:00:08,220 --> 00:00:14,550 So we have basically this youth case in which I have to predict what can be the weight of a person if 5 00:00:14,550 --> 00:00:16,210 that person is dismissed. 6 00:00:16,560 --> 00:00:22,230 So what we have to do, basically, we have to simply put, we have to simply consider this as my ex. 7 00:00:22,560 --> 00:00:23,550 So here it is nothing. 8 00:00:23,550 --> 00:00:27,600 But it is just my it just my 1:00 a.m. policy. 9 00:00:27,780 --> 00:00:35,360 And basically, I have to find this and I have to find a C simple that's that's followed statement solved. 10 00:00:35,490 --> 00:00:36,920 That's what we are trying to learn. 11 00:00:37,260 --> 00:00:40,350 So basically for this, you need some basic math. 12 00:00:40,350 --> 00:00:46,410 You need some basics like what is a slope, what is daddy with some basics like cooking. 13 00:00:46,710 --> 00:00:50,670 So that's what we all have covered in the previous session. 14 00:00:51,000 --> 00:00:53,640 So we have learned this is exactly my daddy do. 15 00:00:53,640 --> 00:00:56,640 And what are the practical uses of all these things? 16 00:00:56,640 --> 00:00:59,790 And and you know what what what this line is all about. 17 00:00:59,790 --> 00:01:00,900 And we also learn. 18 00:01:01,140 --> 00:01:01,550 Yeah. 19 00:01:01,560 --> 00:01:07,280 How how to compute the slope between two points and what is a slope at a particular point. 20 00:01:07,710 --> 00:01:16,440 So you will see, let's say, visualize our data and visualize already we we we have this type of line, 21 00:01:16,440 --> 00:01:17,910 we have basically this type of line. 22 00:01:18,270 --> 00:01:23,640 But but it is not necessary that but it is not necessary that it is my best foot line. 23 00:01:23,750 --> 00:01:31,620 I can say it is my best prediction line for my model, because maybe that may be that that according 24 00:01:31,620 --> 00:01:33,780 to data point that this line can be your best. 25 00:01:34,920 --> 00:01:42,570 What either this one or either this one or either this one for how to compute, how to compute ridgeline, 26 00:01:43,290 --> 00:01:48,830 how to compute which line is your best line, which one is the best, how to compute. 27 00:01:49,170 --> 00:01:52,990 Because you see, you've got to have multiple you can have tons of lines in this data. 28 00:01:53,010 --> 00:01:59,160 When you going to 100, you can have a lack of lines for how to compute with the regression line is 29 00:01:59,160 --> 00:02:03,420 your best line, which is the best prediction line? 30 00:02:03,480 --> 00:02:03,910 Hope. 31 00:02:04,230 --> 00:02:06,790 Computer support is what we have to do. 32 00:02:06,810 --> 00:02:09,010 We have simple terminology here. 33 00:02:09,390 --> 00:02:11,760 Let me let me open a new page a little bit. 34 00:02:11,970 --> 00:02:13,740 Yeah, let me open it. 35 00:02:14,280 --> 00:02:16,130 So here what I'm going to do next. 36 00:02:16,250 --> 00:02:17,840 Again, I'm going to visualize this, OK? 37 00:02:18,510 --> 00:02:23,550 Let's say this is exactly my data security. 38 00:02:23,550 --> 00:02:29,130 Don't get me again remove let me not consider this too much data points. 39 00:02:29,190 --> 00:02:31,140 Let me first draw my line. 40 00:02:31,140 --> 00:02:33,740 Let me draw this line so very forth. 41 00:02:33,780 --> 00:02:37,500 I'm just going to say I'm just going to draw this line over here. 42 00:02:38,880 --> 00:02:42,380 So let's say this is my prediction because that is prediction. 43 00:02:42,930 --> 00:02:45,540 Alexa, I have some data points over here. 44 00:02:45,570 --> 00:02:48,960 Let's say here I have some let's say here I have some. 45 00:02:49,170 --> 00:02:55,680 If I were to consider my previous case in which basically I have to predict what can be the weight of 46 00:02:55,680 --> 00:02:57,650 a person who had this much height. 47 00:02:58,020 --> 00:02:59,850 So let me just consider that. 48 00:03:01,420 --> 00:03:09,880 Here I had basically my height and I had basically my weight over there, and I have to tread this on 49 00:03:09,880 --> 00:03:15,540 a basis of whatever height, whatever input I'm going to give to this production line. 50 00:03:16,090 --> 00:03:18,520 So let's say these are all the data projected. 51 00:03:18,540 --> 00:03:19,750 These are one data point. 52 00:03:20,050 --> 00:03:21,640 These are all the data points. 53 00:03:22,790 --> 00:03:29,690 And let's say this is the final line that I have achieved over here for what you have to do, so you 54 00:03:29,690 --> 00:03:39,170 have to select you have to select that line as your best line, which has least Atala, which has at 55 00:03:39,180 --> 00:03:39,530 least. 56 00:03:40,310 --> 00:03:44,470 Now you will think how to compute your error a few hours. 57 00:03:44,490 --> 00:03:44,870 Nothing. 58 00:03:44,870 --> 00:03:47,350 But this is your Adam. 59 00:03:49,130 --> 00:03:50,960 Similarly, this is your ATM. 60 00:03:51,560 --> 00:03:54,710 This is your ATM, this is your ATM. 61 00:03:55,730 --> 00:03:56,760 This is your ATM. 62 00:03:56,780 --> 00:03:57,890 This is your ATM. 63 00:03:58,520 --> 00:03:59,680 This is your ATM. 64 00:04:00,200 --> 00:04:01,370 So go to our line. 65 00:04:01,520 --> 00:04:05,340 Has at least atter that will get selected as Pascaline. 66 00:04:05,750 --> 00:04:07,540 That's what that's what I am trying to achieve. 67 00:04:08,330 --> 00:04:15,800 So let me let me write a very simple, simple example of the very first time with the celebrities. 68 00:04:15,810 --> 00:04:16,130 It is. 69 00:04:16,380 --> 00:04:18,430 Yeah, that's it. 70 00:04:18,480 --> 00:04:21,730 Oh, let me let me go see the very simple wizard. 71 00:04:21,980 --> 00:04:27,290 Let's say this is my let's say this is my I can't see this is my actual data point. 72 00:04:28,100 --> 00:04:33,700 And let's say if I were to consider with respect to the spiritual aspect of this spiritual life. 73 00:04:34,010 --> 00:04:38,320 So let's say this is my let's listen to and here's something I say. 74 00:04:38,330 --> 00:04:40,790 My prediction happens. 75 00:04:41,570 --> 00:04:49,340 So whatever is a difference between this, what is the difference between this is exactly my ATM is 76 00:04:49,340 --> 00:04:50,450 exactly my ATM. 77 00:04:50,720 --> 00:04:59,060 Let's say in actual your weight is eighty eight and but your line is predicting that your weight is 78 00:04:59,060 --> 00:05:00,110 70 80. 79 00:05:00,390 --> 00:05:03,350 It means you have Aronoff 18 Gidgee. 80 00:05:04,280 --> 00:05:05,600 That's what I'm trying to say. 81 00:05:05,870 --> 00:05:07,520 You I'm trying to make you understand. 82 00:05:08,840 --> 00:05:14,330 So what you have to do, you have to minimize you have to minimize this error. 83 00:05:14,720 --> 00:05:19,460 You have to minimize this at the journey that you will achieve your best. 84 00:05:19,460 --> 00:05:25,520 Footlight, depending upon what you face, you have for how to achieve bastardly. 85 00:05:25,610 --> 00:05:29,600 So it's a simple, simple mathematical equation or it is nothing. 86 00:05:29,600 --> 00:05:32,510 But what is my actual prediction? 87 00:05:32,750 --> 00:05:38,000 I can see what is my actual data and what I'm going to predict. 88 00:05:38,660 --> 00:05:43,700 So there's nothing but is square of this now you will think what I'm going to consider is graph this 89 00:05:43,700 --> 00:05:44,270 very first. 90 00:05:44,300 --> 00:05:45,590 Let me tell you a scenario. 91 00:05:45,800 --> 00:05:49,320 So very first you have to compute the entire atom. 92 00:05:49,490 --> 00:05:51,100 So what exactly is data out of? 93 00:05:51,110 --> 00:05:52,640 Let's say this is my data point one. 94 00:05:52,640 --> 00:05:53,240 This is two. 95 00:05:53,240 --> 00:05:53,920 This is three. 96 00:05:53,930 --> 00:05:54,650 This is four. 97 00:05:55,040 --> 00:05:56,110 And this is five. 98 00:05:56,120 --> 00:05:56,810 This is six. 99 00:05:56,810 --> 00:05:57,530 This is seven. 100 00:05:57,560 --> 00:06:02,030 So it is nothing but some additional I close to one to seven. 101 00:06:02,420 --> 00:06:05,750 This why I still you can consider it is why as well. 102 00:06:06,170 --> 00:06:08,680 You can say it is why and it is why I buy. 103 00:06:09,110 --> 00:06:11,510 So now you will think why I have a cloud over here. 104 00:06:11,510 --> 00:06:17,290 Because you will see here you have whatever you have postulated and here you have negative at it. 105 00:06:17,660 --> 00:06:21,200 So if you really do that summation, maybe it will cancel that. 106 00:06:21,470 --> 00:06:24,590 It makes no sense at all of your error. 107 00:06:24,920 --> 00:06:27,290 Therefore I would put it that way. 108 00:06:27,500 --> 00:06:33,950 So now basically you have to compute something known as your mune is good atah, which is nothing but 109 00:06:34,400 --> 00:06:38,390 one by act, which is nothing but my total number of data points. 110 00:06:39,680 --> 00:06:46,000 To turn total number of the tabloids and and and hear what you have computed, I called two one until 111 00:06:46,010 --> 00:06:56,490 seven and why minus y I all you can get is that it is nothing but my prediction value or because the 112 00:06:56,510 --> 00:07:00,150 prediction is in such you guessed it is it is great. 113 00:07:00,220 --> 00:07:02,840 It is your prediction and it is your actual. 114 00:07:03,230 --> 00:07:05,520 It is you actually. 115 00:07:05,840 --> 00:07:08,290 So who do ever lie. 116 00:07:08,570 --> 00:07:20,780 Has this list mse whosever line has this least masc that line, that line gets selected as my best line, 117 00:07:20,840 --> 00:07:26,200 best fit regression line for how to select this best foot regression line. 118 00:07:26,420 --> 00:07:28,420 Simple that has EMC. 119 00:07:28,640 --> 00:07:35,390 So now how to find this masc or you can see how to find your best regression and you will see here you 120 00:07:35,390 --> 00:07:36,770 have multiple regression. 121 00:07:37,010 --> 00:07:39,300 It means you have multiple matches as well. 122 00:07:39,680 --> 00:07:41,870 Let's say you have something. 123 00:07:41,870 --> 00:07:42,740 MSU one. 124 00:07:43,430 --> 00:07:44,270 MASC two. 125 00:07:46,100 --> 00:07:46,500 Three. 126 00:07:48,820 --> 00:07:55,150 They'll tell you have hundred of a little meniscal, Adam, with respect to this, will have some your 127 00:07:55,660 --> 00:07:59,930 slope and your your your constant let me go through the ceiling. 128 00:08:00,160 --> 00:08:02,510 Similarly, you have to come whether. 129 00:08:02,950 --> 00:08:03,670 Because you will see. 130 00:08:04,060 --> 00:08:04,610 You'll see. 131 00:08:04,630 --> 00:08:07,900 Let me let me open a previous U.S. exit with respect to this line. 132 00:08:07,900 --> 00:08:11,940 Let's say with respect to this purple line, let's say it's messy. 133 00:08:11,950 --> 00:08:13,450 That's get added. 134 00:08:13,730 --> 00:08:14,800 That is one hundred. 135 00:08:15,160 --> 00:08:20,710 And let's say it has slope of let's say point eight and exit as content of exit four. 136 00:08:20,950 --> 00:08:22,710 So here you have another beer. 137 00:08:22,840 --> 00:08:25,900 Similarly, with respect to each and every line, you have some pair. 138 00:08:26,890 --> 00:08:32,470 So let me let me let me do I need to know every single this this will give something similar like this 139 00:08:32,470 --> 00:08:33,250 will be something. 140 00:08:34,130 --> 00:08:38,860 So if I'm going to plot if I'm going to plot all these things. 141 00:08:39,010 --> 00:08:47,860 So it means definitely I need some kind of three dimensional visual or some kind of three dimensional 142 00:08:47,860 --> 00:08:51,040 thought to understand this, to understand the scenario. 143 00:08:51,910 --> 00:08:53,590 So let me open this plot. 144 00:08:53,770 --> 00:08:54,560 Let me open it. 145 00:08:54,900 --> 00:09:03,160 Yeah, this is exactly this is exactly three dimensional visual where you have something known as slope, 146 00:09:03,340 --> 00:09:04,850 where you have something on a slope. 147 00:09:05,170 --> 00:09:06,880 This is exactly the intercept. 148 00:09:06,910 --> 00:09:13,650 You can either represent it a season and this is this is your square, which is nothing, which is, 149 00:09:13,650 --> 00:09:17,590 I think, your cost function, which you have to be, which you have to optimize. 150 00:09:17,860 --> 00:09:26,200 Would you have to minimize because whoever line, whosever line has a least MSE has a least MSI, that 151 00:09:26,200 --> 00:09:28,410 line gets selected as my best line. 152 00:09:28,480 --> 00:09:30,280 So that's all about the session. 153 00:09:30,490 --> 00:09:31,740 In the upcoming session. 154 00:09:31,750 --> 00:09:37,870 We are going to learn about this plot, what this plot is all about, and, you know, our very famous 155 00:09:37,870 --> 00:09:42,710 approach, which is exactly gradient descent approach as well. 156 00:09:43,090 --> 00:09:44,240 So that's all about it. 157 00:09:44,260 --> 00:09:44,970 Thank you. 158 00:09:45,010 --> 00:09:45,910 Have a nice day. 159 00:09:46,150 --> 00:09:47,050 Keep learning. 160 00:09:47,050 --> 00:09:47,970 Keep growing. 161 00:09:48,520 --> 00:09:49,450 Keep practicing.