1 00:00:03,300 --> 00:00:12,420 Suppose we have some samples, issues of our independent variable X and this one is over dependent variable 2 00:00:12,420 --> 00:00:12,750 Y. 3 00:00:13,410 --> 00:00:16,010 And here are our samples. 4 00:00:16,170 --> 00:00:24,070 So I'm looking for a way to feed a function to satisfy and to cover all those data. 5 00:00:24,270 --> 00:00:27,430 Here I can already try. 6 00:00:28,380 --> 00:00:38,580 OK, not a very good fit, but however, as you can see, this is a parabolic curve for a parabolic 7 00:00:38,580 --> 00:00:39,050 curve. 8 00:00:39,060 --> 00:00:49,770 We can already write a function not very accurate, but here is a zero plus a one X plus a two. 9 00:00:51,480 --> 00:01:01,500 And X Square, we are looking for a tool to give us the best fit today's samples, we are looking for 10 00:01:01,500 --> 00:01:09,300 minimizing the error, the distance between these simple and this curve, it supposedly must be on the 11 00:01:09,300 --> 00:01:09,770 curve. 12 00:01:09,780 --> 00:01:16,350 But however, we cannot just draw a line like that to satisfy all the samples. 13 00:01:16,500 --> 00:01:24,960 We have to just draw a curve to satisfy most of the samples, but we really can't do it for all. 14 00:01:25,200 --> 00:01:27,070 Let me just raise it again. 15 00:01:27,390 --> 00:01:31,890 Let's just try to fit another parabolic curve. 16 00:01:32,850 --> 00:01:34,500 I'm going to say this one maybe. 17 00:01:37,390 --> 00:01:43,150 So we want to use artificial neural network to help us feed this curve. 18 00:01:43,300 --> 00:01:49,300 Now let's check out the error that we can have here, the distance between these two here. 19 00:01:49,790 --> 00:01:55,000 I can call this one E of eye disease. 20 00:01:55,010 --> 00:02:05,570 Or if I and if you just started here, this is our data and this is just what we were able to fit. 21 00:02:06,730 --> 00:02:16,600 This is over ViiV I and this one is white hat to I, the one which we were able to fit of. 22 00:02:16,600 --> 00:02:24,250 Our task here is to minimize the error by using the meniscus or they're looking to minimizing these 23 00:02:24,250 --> 00:02:24,880 errors. 24 00:02:25,150 --> 00:02:26,380 Here we have another one. 25 00:02:26,680 --> 00:02:36,820 This is another EFI and we are looking to minimize these errors and fit the best curve to satisfy all 26 00:02:36,820 --> 00:02:40,330 the data in the technical section. 27 00:02:40,600 --> 00:02:41,260 We all did it. 28 00:02:41,260 --> 00:02:47,200 Learn how to calculate the mean square error and Verity's formula is coming from now. 29 00:02:47,200 --> 00:02:49,340 Let's take a look at this point. 30 00:02:49,960 --> 00:02:56,070 Our goal is to just say that line should be equal to we have. 31 00:02:56,120 --> 00:03:01,210 This is the best situation, which I already explained, that it's not going to happen. 32 00:03:01,220 --> 00:03:07,840 The error cannot be zero because of several factors, especially with the natural systems, both of 33 00:03:07,840 --> 00:03:11,100 them, we are looking to carve a parabolic curve. 34 00:03:11,110 --> 00:03:17,130 So here is a zero plus a one X and a two X square. 35 00:03:17,920 --> 00:03:25,510 The factors that we can change in order to fit the best curve are a zero A1 A2. 36 00:03:25,900 --> 00:03:33,970 So let's find out the best value that we can set for these variables in order to have the best foetid 37 00:03:33,970 --> 00:03:39,330 curve to start working with Matlab tools, we need to have some data. 38 00:03:39,520 --> 00:03:41,320 So let's generate a data X. 39 00:03:41,320 --> 00:03:49,230 If I am going to say it's between zero to two pi y if I equals two signers of X of mine. 40 00:03:49,390 --> 00:03:54,660 This is just one sample to see how the neural network is going to work. 41 00:03:55,540 --> 00:04:03,550 And of course I is going from one to and in the other hand we have over process, so we are feting X, 42 00:04:03,550 --> 00:04:06,220 which is over input to two systems. 43 00:04:06,520 --> 00:04:13,400 One is of a process which is a sign of X and the other one is artificial neural network. 44 00:04:13,420 --> 00:04:22,090 So our goal is to train this network by showing it several samples and we want to have the same output, 45 00:04:22,120 --> 00:04:27,880 or at least we are trying to minimize the error for these two outputs. 46 00:04:28,210 --> 00:04:32,620 Now let's back to Matlab and generate these samples.