1 00:00:00,970 --> 00:00:02,530 To train the model. 2 00:00:03,040 --> 00:00:06,040 What we will do is we will. 3 00:00:07,200 --> 00:00:08,670 It's quite simple. 4 00:00:08,670 --> 00:00:10,710 Just let me just take some space here. 5 00:00:11,810 --> 00:00:14,840 And it will be modal equals. 6 00:00:16,530 --> 00:00:18,990 Dee dee dee dee dee. 7 00:00:19,020 --> 00:00:19,470 Sorry. 8 00:00:19,500 --> 00:00:23,250 Will deep dot model. 9 00:00:23,280 --> 00:00:24,030 Model. 10 00:00:24,630 --> 00:00:26,610 And we will pass two things. 11 00:00:26,610 --> 00:00:30,780 The first one is the data we have, which is this one. 12 00:00:33,210 --> 00:00:36,330 And the second thing we will pass is the neural network. 13 00:00:36,330 --> 00:00:37,440 So network. 14 00:00:42,480 --> 00:00:43,300 And that's it. 15 00:00:43,320 --> 00:00:44,280 This is the model. 16 00:00:44,550 --> 00:00:47,070 Now how to train it. 17 00:00:47,100 --> 00:00:49,650 We will do also the. 18 00:00:50,860 --> 00:00:52,360 Model dot. 19 00:00:53,790 --> 00:00:54,540 Compile. 20 00:00:55,930 --> 00:00:57,430 And we're using Adam. 21 00:00:59,500 --> 00:01:03,910 With with a learning rate of. 22 00:01:05,030 --> 00:01:07,040 E to the power minus three. 23 00:01:08,150 --> 00:01:11,900 And now we will compile it. 24 00:01:13,530 --> 00:01:17,210 And we will start the training soon. 25 00:01:17,220 --> 00:01:22,350 What we will do or let me start the training and then I will explain more history. 26 00:01:25,810 --> 00:01:27,880 His story. 27 00:01:29,370 --> 00:01:30,420 And train. 28 00:01:31,960 --> 00:01:36,460 State equals model dot train. 29 00:01:38,580 --> 00:01:42,870 Iterations equals 15. 30 00:01:43,680 --> 00:01:44,330 Thousand. 31 00:01:46,060 --> 00:01:46,790 And reasons. 32 00:01:46,810 --> 00:01:47,140 Yeah. 33 00:01:49,050 --> 00:01:50,230 Loss. 34 00:01:50,520 --> 00:01:53,580 His two seems okay and shift enter. 35 00:01:54,820 --> 00:01:59,770 So now the model is actually training and, um. 36 00:02:01,600 --> 00:02:02,590 We can see. 37 00:02:02,590 --> 00:02:04,270 It's like. 38 00:02:06,360 --> 00:02:11,220 But I think I wanted to print this thing. 39 00:02:12,680 --> 00:02:14,030 Okay, Just let me. 40 00:02:15,920 --> 00:02:16,430 Stop it. 41 00:02:16,490 --> 00:02:17,900 Think it didn't run the whole. 42 00:02:17,990 --> 00:02:19,370 Let's run it from the beginning. 43 00:02:20,580 --> 00:02:22,470 Why didn't he print these things? 44 00:02:22,470 --> 00:02:24,540 I think it we didn't shift into it. 45 00:02:29,550 --> 00:02:30,000 Yeah. 46 00:02:30,030 --> 00:02:30,360 Okay. 47 00:02:30,360 --> 00:02:31,310 Now we can see them. 48 00:02:31,320 --> 00:02:32,040 We didn't. 49 00:02:32,070 --> 00:02:34,170 We wrote them, but we didn't shift enter. 50 00:02:34,170 --> 00:02:41,790 We didn't put them inside the, the, the ram didn't upload it so input or the actually the, the whole 51 00:02:41,790 --> 00:02:42,210 thing. 52 00:02:42,330 --> 00:02:42,630 Yeah. 53 00:02:42,720 --> 00:02:43,350 Mainly the ram. 54 00:02:43,680 --> 00:02:49,590 So the input we have this is the actual input that goes just a little bit. 55 00:02:49,620 --> 00:02:52,860 We talk about this one the a. 56 00:02:54,600 --> 00:02:59,940 In this issue how to change the lambda function, because I think it's a little bit confusing. 57 00:03:00,060 --> 00:03:06,540 So we talked about we have an input array and we need to see the shape. 58 00:03:06,540 --> 00:03:12,870 The input array is simply going to be the boundary point or in the in case of initial condition, it 59 00:03:12,870 --> 00:03:21,480 will be the initial condition points and it will be given in many different time steps, of course. 60 00:03:21,570 --> 00:03:28,140 So input array will be 80 to 80 rows and two columns because we have. 61 00:03:29,880 --> 00:03:34,320 This one, which is basically zero and the time. 62 00:03:35,120 --> 00:03:35,900 And here. 63 00:03:35,900 --> 00:03:38,110 What if we do 0 to 1? 64 00:03:38,120 --> 00:03:39,910 It will print it this way. 65 00:03:39,920 --> 00:03:43,130 1001111 this way. 66 00:03:43,220 --> 00:03:46,520 And if we change this to. 67 00:03:47,150 --> 00:03:55,640 Well, we put it like if we didn't put 0 to 1, what will happen is like. 68 00:03:56,260 --> 00:04:03,760 If we put it this way, the shape will be only 80 and it will look like this array, which is this is 69 00:04:03,760 --> 00:04:04,840 not what we want. 70 00:04:05,170 --> 00:04:11,650 So we have to be careful about the the values that we need to consider. 71 00:04:12,960 --> 00:04:19,500 Anyway, it will continue to train and he passed two times, one for every boundary condition. 72 00:04:19,500 --> 00:04:21,960 So he passed first time and he passed second time. 73 00:04:21,960 --> 00:04:23,900 This is why he printed two times. 74 00:04:23,910 --> 00:04:27,810 And basically this is finished the training itself. 75 00:04:28,760 --> 00:04:31,130 The next step is going to be. 76 00:04:32,120 --> 00:04:34,160 Well, basically the. 77 00:04:35,830 --> 00:04:40,780 A the that we need to further improve the accuracy. 78 00:04:40,780 --> 00:04:43,600 So model dot compile. 79 00:04:44,790 --> 00:04:46,020 Compile. 80 00:04:47,350 --> 00:04:53,440 And we will use a slash bfgs. 81 00:04:54,470 --> 00:04:57,140 A B method, which is a method. 82 00:04:57,140 --> 00:04:58,610 It's a limited memory. 83 00:04:58,610 --> 00:05:01,250 Something it's it's it's related to. 84 00:05:01,280 --> 00:05:05,840 It will give us a good accuracy in, in in kind of limited memory. 85 00:05:05,960 --> 00:05:12,590 So we can of course, we can search about this, how it works more, but mainly we use it in pens for 86 00:05:12,590 --> 00:05:15,010 giving getting a better accuracy. 87 00:05:15,020 --> 00:05:17,420 It's not like Adam will not work. 88 00:05:17,450 --> 00:05:18,350 It will work. 89 00:05:18,350 --> 00:05:21,530 But this is a common practice to use this one. 90 00:05:21,950 --> 00:05:30,890 So model again train we need now we we we train again the model, but this time we will not define how 91 00:05:30,890 --> 00:05:37,340 many iterations it will until it will converge and then shift enter. 92 00:05:38,120 --> 00:05:43,880 So again, he's took the thing and he will now start the training. 93 00:05:44,850 --> 00:05:47,550 A after it will finish it. 94 00:05:47,580 --> 00:05:56,880 Hopefully in the next class we will see how these models, the results generated by this model and we 95 00:05:56,880 --> 00:05:59,730 will explain what's happening in these results. 96 00:05:59,730 --> 00:06:01,950 So see you in next class.