1 00:00:00,360 --> 00:00:00,720 Hello. 2 00:00:00,750 --> 00:00:01,380 Welcome back. 3 00:00:01,380 --> 00:00:02,070 Let's continue. 4 00:00:02,100 --> 00:00:07,830 So we're going to write the function for learning or optimizing as some would call it. 5 00:00:07,830 --> 00:00:11,530 I'm going to save this and continue from a new one. 6 00:00:11,550 --> 00:00:19,270 Call this number five and then I'm going to come down here and create a new function called Len. 7 00:00:19,730 --> 00:00:27,620 I said def Len and this function is going to okay this word is taking 8 00:00:32,830 --> 00:00:34,000 I'll say and Len 9 00:00:37,590 --> 00:00:38,940 Oh no sorry. 10 00:00:39,280 --> 00:00:40,120 We can use Len. 11 00:00:40,150 --> 00:00:40,410 Yeah. 12 00:00:40,630 --> 00:00:42,520 Sorry about that because he was blue. 13 00:00:42,520 --> 00:00:45,060 I thought it was a reserved key word in python. 14 00:00:45,100 --> 00:00:46,050 But it's not. 15 00:00:46,610 --> 00:00:46,980 Okay. 16 00:00:47,020 --> 00:00:52,450 So he's going to take a number of arguments we're going to pass that weight the buyers the training 17 00:00:52,480 --> 00:01:01,240 data the labels we're going to pass number of iterations as well. 18 00:01:01,240 --> 00:01:03,850 And then we can pass the learning rate 19 00:01:06,800 --> 00:01:08,620 yes and then that's fine. 20 00:01:09,340 --> 00:01:16,810 So yeah I'm going to create a list here an empty list called cost 21 00:01:20,160 --> 00:01:27,280 and then I can open them the loop for learning I'll see for i in range. 22 00:01:29,020 --> 00:01:29,950 She can guess 23 00:01:32,900 --> 00:01:34,250 number of iterations 24 00:01:46,900 --> 00:01:52,810 then we can call our product to function propagation over here or call this. 25 00:01:53,470 --> 00:01:57,600 And then remember propagation is going to return the gradients and the cost. 26 00:01:58,210 --> 00:02:08,060 So you can't simply come here and see gradients Cost. 27 00:02:08,530 --> 00:02:10,760 Cost is different from costs. 28 00:02:10,830 --> 00:02:22,210 Cost this plural to singular cause propagation then I'm going to pass the parameters required W B X 29 00:02:22,330 --> 00:02:24,200 and Y like this. 30 00:02:24,760 --> 00:02:32,150 And then once we've computed a cost for so remember each example we compute that cost and then we add 31 00:02:32,180 --> 00:02:32,700 it here. 32 00:02:33,280 --> 00:02:37,920 So I'm going to append this to our costs list. 33 00:02:37,930 --> 00:02:40,840 This costs costs plural. 34 00:02:41,020 --> 00:02:46,830 I'm going to append costs that append this singular cost like this. 35 00:02:47,680 --> 00:02:48,540 Okay. 36 00:02:48,730 --> 00:02:52,820 Once that is done we have to update the wait. 37 00:02:52,830 --> 00:02:57,430 I'll come here and see it W equals W minus. 38 00:02:57,670 --> 00:02:59,140 We apply the lending rate 39 00:03:02,200 --> 00:03:07,780 multiplied by T W and then we update the bias as well. 40 00:03:07,870 --> 00:03:13,790 B equals B minus. 41 00:03:13,990 --> 00:03:15,370 We apply the Lenin rate 42 00:03:19,800 --> 00:03:23,610 and then be like this. 43 00:03:24,090 --> 00:03:24,510 Okay. 44 00:03:27,820 --> 00:03:29,110 So we can print the cost. 45 00:03:29,110 --> 00:03:39,970 We can say if it if it runs for let's say 50 times print that cost so I can see if if by percentage 46 00:03:40,230 --> 00:03:47,050 hobo there's a percentage 50 because you then I'm going to print the cost print 47 00:03:53,630 --> 00:03:54,110 stuff. 48 00:03:54,130 --> 00:04:00,250 Iteration percentage I don't percentage F 49 00:04:06,290 --> 00:04:12,880 and I'm simply going to pass the value here and the cost here. 50 00:04:12,880 --> 00:04:14,760 Like this. 51 00:04:14,760 --> 00:04:15,160 Right 52 00:04:20,500 --> 00:04:27,730 and um we can store the parameters in the key valued pair in a dictionary and then return the parameters 53 00:04:27,820 --> 00:04:34,080 as well as the gradient after Len and after the optimization process. 54 00:04:34,180 --> 00:04:43,060 So I'm going to create paradigms here and store the WNBA values course we need then we would need them 55 00:04:43,060 --> 00:04:46,370 later firearms. 56 00:04:46,510 --> 00:04:59,520 And then I'll say w w and then B B right. 57 00:04:59,620 --> 00:05:02,140 And then the gradient reds 58 00:05:06,520 --> 00:05:10,480 Oh c d w w 59 00:05:13,330 --> 00:05:29,120 DP DP and I'm going to return parameters and gradients return but Rams reds and also costs. 60 00:05:29,120 --> 00:05:30,960 Right. 61 00:05:31,740 --> 00:05:33,450 Okay. 62 00:05:34,330 --> 00:05:35,410 So this is it. 63 00:05:35,410 --> 00:05:42,670 This is our function to learn the next lesson we going to write the function to help us predict. 64 00:05:43,560 --> 00:05:44,570 Yeah. 65 00:05:44,890 --> 00:05:47,700 So this order is for this lesson if you have any questions just let me know. 66 00:05:47,710 --> 00:05:48,490 I'll see you later.