WEBVTT 00:00.420 --> 00:02.880 -: Hello and welcome to this Python tutorial. 00:02.880 --> 00:04.620 All right, it's a very quick tutorial today 00:04.620 --> 00:06.690 to make the score function, 00:06.690 --> 00:10.080 and so basically, this function will just compute the score 00:10.080 --> 00:12.840 on the sliding window of the rewards, 00:12.840 --> 00:15.960 and so basically, we will very simply compute the mean 00:15.960 --> 00:18.180 of all the rewards in the reward window. 00:18.180 --> 00:19.650 So this will be very simple. 00:19.650 --> 00:20.880 Let's do this now. 00:20.880 --> 00:22.500 We're gonna make this new function 00:22.500 --> 00:24.480 that we're gonna call score. 00:24.480 --> 00:27.270 And the score function will just take the argument self 00:27.270 --> 00:29.850 because basically we don't need anything. 00:29.850 --> 00:30.870 We need to take self 00:30.870 --> 00:33.810 because, of course, we will take self.rewardWindow. 00:33.810 --> 00:35.160 So just self. 00:35.160 --> 00:37.800 And then colon and there we go. 00:37.800 --> 00:39.990 It's gonna take one line of code. 00:39.990 --> 00:41.850 So we wanna compute the mean 00:41.850 --> 00:43.830 of all the rewards in the reward window. 00:43.830 --> 00:46.710 So that's basically the sum of all the rewards 00:46.710 --> 00:48.120 in this reward window 00:48.120 --> 00:50.250 that are between -1 and +1 00:50.250 --> 00:54.150 divided by the total number of elements in this window. 00:54.150 --> 00:55.170 So let's do this. 00:55.170 --> 00:57.270 We're directly going to return that. 00:57.270 --> 00:59.730 So I'm starting with return. 00:59.730 --> 01:01.800 And so we need to take the sum 01:01.800 --> 01:05.220 of all the rewards in the reward window, 01:05.220 --> 01:06.570 and to do this, we simply need 01:06.570 --> 01:08.820 to take the reward window itself 01:08.820 --> 01:11.823 and so I'm inputting here self.rewardWindow. 01:14.542 --> 01:15.840 And so very simply, 01:15.840 --> 01:20.790 this will sum all the elements inside rewardWindow. 01:20.790 --> 01:22.200 So that's pretty practical. 01:22.200 --> 01:23.910 And then to get the mean, 01:23.910 --> 01:26.730 we need to divide this sum 01:26.730 --> 01:29.490 by the number of elements in the rewardWindow, 01:29.490 --> 01:31.170 and to get the number of elements, 01:31.170 --> 01:34.200 well, we need to take the len function, 01:34.200 --> 01:37.023 and then we take our rewardWindow again. 01:39.210 --> 01:40.043 There it is. 01:40.043 --> 01:42.810 But now we just need to be careful with something. 01:42.810 --> 01:47.190 It's that len self.rewardWindow is a denominator 01:47.190 --> 01:50.160 and this must absolutely not be equal to zero. 01:50.160 --> 01:52.020 No matter what, we need to avoid this 01:52.020 --> 01:54.360 and to make sure that the denominator here 01:54.360 --> 01:55.770 is not equal to zero. 01:55.770 --> 01:58.800 We're gonna add this safety trick. 01:58.800 --> 02:01.510 We're gonna add here a plus one 02:02.460 --> 02:05.670 so that len self.rewardWindow plus one 02:05.670 --> 02:07.350 will never be equal to zero. 02:07.350 --> 02:09.450 If the denominator here is equal to zero, 02:09.450 --> 02:11.370 this will crash your system. 02:11.370 --> 02:13.200 So we must avoid it, 02:13.200 --> 02:15.000 and that's totally fine to add a plus one. 02:15.000 --> 02:17.850 We will still get a good measure in the score. 02:17.850 --> 02:18.780 All right, perfect. 02:18.780 --> 02:20.010 And so that's all. 02:20.010 --> 02:21.480 We have our score function, 02:21.480 --> 02:23.640 which will give us the mean of the rewards 02:23.640 --> 02:25.200 in the signing window. 02:25.200 --> 02:28.050 All right, now let's move on to the next function, 02:28.050 --> 02:29.520 which is the save function 02:29.520 --> 02:31.530 that will save your model, 02:31.530 --> 02:33.480 that is save the brain of your core 02:33.480 --> 02:34.830 so that you can then be able 02:34.830 --> 02:37.530 to reuse it by loading it with another function 02:37.530 --> 02:39.840 that we'll make after the save function. 02:39.840 --> 02:43.440 So that's really practical to have this save trick, 02:43.440 --> 02:45.510 save function to save your models 02:45.510 --> 02:47.130 in case you want to reuse them 02:47.130 --> 02:49.590 for any kind of purpose where they can be useful. 02:49.590 --> 02:51.540 So that's what we'll do in the next tutorial. 02:51.540 --> 02:53.373 And until then, enjoy AI.