1 00:00:00,750 --> 00:00:01,980 Hello, welcome back. 2 00:00:02,220 --> 00:00:09,210 In into session, we are going to practice with a different data set to explore more about math functions 3 00:00:09,480 --> 00:00:14,980 and to see what can we do if we have more than one outputs for this purpose. 4 00:00:15,000 --> 00:00:23,190 I'm going to just use one of the datasets that Matlab is providing us to go to the app and open the 5 00:00:23,190 --> 00:00:25,220 neural net fitting tools. 6 00:00:25,590 --> 00:00:29,230 Click up next here we can see different data set. 7 00:00:29,520 --> 00:00:30,810 Let's take a look at them. 8 00:00:31,450 --> 00:00:33,750 We have a simple fitting problem. 9 00:00:34,140 --> 00:00:37,620 We have a body fat problem in this example. 10 00:00:37,620 --> 00:00:43,050 We have 13 different inputs, total of two hundred fifty two samples. 11 00:00:43,440 --> 00:00:50,670 These inputs are age, weight, height and the information about the body part. 12 00:00:51,120 --> 00:00:55,740 And as an adult food, it just gave us a body fat percentage. 13 00:00:56,400 --> 00:00:59,810 Another example that we have here is building analogy. 14 00:01:00,330 --> 00:01:08,160 This is a data set of four thousand two hundred eight samples with fourteen inputs and three outputs. 15 00:01:08,850 --> 00:01:11,640 And here we have chemical. 16 00:01:11,640 --> 00:01:12,660 We have a gene. 17 00:01:13,030 --> 00:01:16,290 Let me check this engine data set problem. 18 00:01:17,760 --> 00:01:22,980 In this example, we have two inputs which are the feel right and a speed. 19 00:01:23,280 --> 00:01:32,400 And we have two engine targets or output which are torque and nitrous oxide emissions. 20 00:01:33,360 --> 00:01:38,490 We have total of one thousand one hundred ninety nine samples of H. 21 00:01:39,090 --> 00:01:41,250 This sounds good for our purpose. 22 00:01:41,250 --> 00:01:45,690 So I will load this example, but not from here. 23 00:01:46,170 --> 00:01:47,610 Let me check the filename. 24 00:01:47,940 --> 00:01:50,220 This one is engine data dataset. 25 00:01:50,670 --> 00:01:53,900 I'm going to load it here in my command life. 26 00:01:53,910 --> 00:01:57,450 Just copy this name and close this one degrees. 27 00:01:57,570 --> 00:02:04,320 Then I'm not going to load another neural network is because we have already created an advanced script 28 00:02:04,590 --> 00:02:07,610 which gives us more functions. 29 00:02:07,620 --> 00:02:09,540 So I want to use the same. 30 00:02:09,540 --> 00:02:17,730 And I don't want to create another advanced script from the scratch here and the command window to see 31 00:02:17,730 --> 00:02:18,390 the data. 32 00:02:18,390 --> 00:02:26,310 You can simply try to load the name of your dataset, as you can see in a word, caspase. 33 00:02:26,310 --> 00:02:30,810 I already have engine targets and I have engine inputs. 34 00:02:32,190 --> 00:02:34,020 But let's do something else. 35 00:02:36,520 --> 00:02:46,810 I can create a variable named data and I can load my samples, the data said, in these data so I can 36 00:02:46,810 --> 00:02:49,240 manage it better in a day. 37 00:02:49,480 --> 00:02:55,750 We have engine inputs, which are two different inputs and engine targets, which are, again, two 38 00:02:55,750 --> 00:02:57,270 different outputs. 39 00:02:58,240 --> 00:03:06,790 Let's create a new one and copy and paste all this creep that we created in advance, a script number 40 00:03:06,790 --> 00:03:14,450 two and just pasted to here and before doing anything else, save over data and give it a name. 41 00:03:14,920 --> 00:03:18,850 This time I'm going to call it Advancer script number tweet. 42 00:03:18,850 --> 00:03:24,220 We underscore gene underscore dataset. 43 00:03:26,150 --> 00:03:29,310 OK, well, then I'll settle to start here. 44 00:03:30,350 --> 00:03:38,270 We really don't need these inputs and targets, we are going to use their engine data set that's Maslov 45 00:03:38,270 --> 00:03:39,470 is providing for us. 46 00:03:39,470 --> 00:03:45,830 If you have a different form, you can just employ that in Matlab and you can use your own data set, 47 00:03:46,070 --> 00:03:49,370 which usually using Excel, fine, you can import it here. 48 00:03:49,850 --> 00:03:53,930 First thing to do, create a data and then load. 49 00:03:57,030 --> 00:03:58,230 Engine data set. 50 00:04:02,780 --> 00:04:10,960 Next one is I want my pooch to being a variable named X, X is the name that I use for the entire program, 51 00:04:11,210 --> 00:04:12,850 so I'm not going to change it here. 52 00:04:12,860 --> 00:04:14,450 Let's just call them X. 53 00:04:14,810 --> 00:04:16,190 What are my inputs? 54 00:04:16,640 --> 00:04:18,710 They are changing inputs. 55 00:04:18,710 --> 00:04:22,340 They called Indurain inputs from data. 56 00:04:22,530 --> 00:04:24,650 Let's call InGen inputs. 57 00:04:24,980 --> 00:04:26,180 I need all of them. 58 00:04:26,180 --> 00:04:28,520 So I'm I'm going to put a semicolon. 59 00:04:28,880 --> 00:04:36,830 And as my target seems to be defined it as a name t before in our advantage, I'm going to use the same 60 00:04:36,830 --> 00:04:44,990 name ts equal from Daytop this time load my targets, which are InGen targets. 61 00:04:45,410 --> 00:04:49,790 But the point here is we have two different outputs. 62 00:04:50,480 --> 00:04:56,400 It means that we need to train as separate neural network for each output. 63 00:04:57,020 --> 00:05:04,970 There are some situations that outputs and targets are having some high correlation and they are related. 64 00:05:05,180 --> 00:05:10,460 So we can just use one neural network to train for the same outputs. 65 00:05:10,880 --> 00:05:19,280 But here we are not sure if the outputs are going to fit and if one neural network with the same number 66 00:05:19,280 --> 00:05:24,020 of player with the same algorithm is going to work for both outputs. 67 00:05:24,020 --> 00:05:30,680 That's why I'm going to first choose the output, number one, and then next time I will create another 68 00:05:30,680 --> 00:05:31,160 advance. 69 00:05:31,160 --> 00:05:34,210 A script called for the second target. 70 00:05:35,900 --> 00:05:39,040 Put this in McAloon and we're all set. 71 00:05:40,190 --> 00:05:46,670 You can already run your program to see if everything is working, though. 72 00:05:48,590 --> 00:05:52,010 OK, Vringo, it's working. 73 00:05:52,010 --> 00:05:56,120 You can see different figures that we created. 74 00:05:56,120 --> 00:05:57,860 You can see the performance. 75 00:05:58,370 --> 00:05:59,180 It's working. 76 00:05:59,180 --> 00:06:03,230 Well, let's do a little bit adjustment here. 77 00:06:04,010 --> 00:06:13,250 I want to have more than one hundred because one hundred just stopped our training and it might not 78 00:06:13,250 --> 00:06:15,950 really be efficient for this example. 79 00:06:15,950 --> 00:06:22,580 And this data set here it is number of folks I will choose five hundred. 80 00:06:23,510 --> 00:06:29,330 And then another team that I'm going to do is to make all these plots as a comment. 81 00:06:29,780 --> 00:06:32,960 So every time that I'm running the program, they won't pop up. 82 00:06:33,800 --> 00:06:37,400 You can use plot three to show it three D plot. 83 00:06:37,730 --> 00:06:40,310 This plot tree has three elements. 84 00:06:40,580 --> 00:06:43,160 We have two inputs and we have one output. 85 00:06:43,170 --> 00:06:46,730 So first, let's show the input. 86 00:06:46,730 --> 00:06:53,060 Number one, the first input, and then let's show the second input. 87 00:06:55,160 --> 00:07:01,400 And finally, I want to see my targets and let's show these targets. 88 00:07:01,400 --> 00:07:03,800 Sweet that. 89 00:07:05,150 --> 00:07:13,310 OK, here we have if you just type weed on, you will see better. 90 00:07:14,450 --> 00:07:16,730 We can see our data. 91 00:07:16,730 --> 00:07:21,560 One of them is my first input, the other one is my second input. 92 00:07:21,800 --> 00:07:24,560 And here I can see the output. 93 00:07:25,610 --> 00:07:31,970 If you just rotated a little bit, we can see somehow they are under one to face. 94 00:07:31,970 --> 00:07:40,640 Each means we can train and know a very good neural network to fit this data and it's good for our approximation 95 00:07:40,640 --> 00:07:41,540 functions. 96 00:07:45,320 --> 00:07:47,670 But let's do something more here. 97 00:07:48,110 --> 00:07:51,770 Just click on Iran one more time to see the feet. 98 00:07:52,580 --> 00:07:56,540 We don't have any feet because we have more than one input. 99 00:07:56,750 --> 00:08:03,070 So we need to have a customized figure to somehow show their feet plot. 100 00:08:05,000 --> 00:08:09,170 Let's go to the plot parts and make some space here. 101 00:08:09,350 --> 00:08:11,890 A start with type figure. 102 00:08:12,050 --> 00:08:17,300 I want to have a figure for this figure I want to plot. 103 00:08:19,110 --> 00:08:29,460 Our troops off of our neural network and show we do the color of black, but hold on, I'm not done 104 00:08:29,460 --> 00:08:30,040 yet. 105 00:08:30,060 --> 00:08:40,440 I will never know their plot to show and same finger the targets we color for it and also make it a 106 00:08:40,440 --> 00:08:41,320 dotted line. 107 00:08:42,750 --> 00:08:50,900 We can also add some commands, use the code Legion, give it a name here. 108 00:08:52,320 --> 00:08:56,400 First one is my neural net outputs. 109 00:08:56,760 --> 00:08:59,880 The second one are my targets. 110 00:09:01,470 --> 00:09:03,900 And finally, give it a title. 111 00:09:04,780 --> 00:09:09,370 Let's just it all day, OK? 112 00:09:09,550 --> 00:09:12,060 I run the program to see the result. 113 00:09:12,060 --> 00:09:13,780 There is an error here. 114 00:09:14,430 --> 00:09:16,160 This one is a semi color. 115 00:09:16,740 --> 00:09:18,210 This one is Leegin. 116 00:09:18,780 --> 00:09:25,200 This is let's hope everything works well this time. 117 00:09:25,260 --> 00:09:27,690 Let's see where we go now. 118 00:09:27,690 --> 00:09:29,400 This is my data. 119 00:09:29,400 --> 00:09:31,170 One of them is Mount Hood. 120 00:09:31,170 --> 00:09:32,910 The other one is the target. 121 00:09:33,240 --> 00:09:43,200 If you just zoom and a part of that, you can see this red dotted line is the targets that we are looking 122 00:09:43,200 --> 00:09:45,120 to fit a neural network. 123 00:09:45,540 --> 00:09:51,000 And this one is a line that my neural network was able to feed, which was a very good fit. 124 00:09:51,000 --> 00:09:52,380 It's almost the same. 125 00:09:54,360 --> 00:09:56,580 So we have a very good neural network.