WEBVTT 00:00.150 --> 00:03.930 Hello again! In this video, we are going to look at random number algorithms. 00:05.370 --> 00:09.720 In fact, there is only one random number algorithm in the library, which is shuffle(). 00:10.470 --> 00:17.850 And this will take a iterator range and put the elements into random order. So we pass it the iterator 00:17.890 --> 00:20.040 range and a random number engine. 00:20.820 --> 00:27.150 And this will do the stuff we saw before: it will use a uniform distribution to get the random numbers. 00:28.320 --> 00:34.230 This will give "perfect shuffling", so-called, which means that all the permutations of the elements are 00:34.230 --> 00:35.130 equally likely. 00:36.600 --> 00:39.240 There was also an algorithm called random_shuffle(). 00:39.870 --> 00:42.260 This used rand() and it is deprecated. 00:42.360 --> 00:44.130 I think actually it has now been removed. 00:48.200 --> 00:57.950 So here is our vector. Here is our random number engine. Here is our call to shuffle(), and here is our shuffled vector. 00:59.090 --> 00:59.410 OK. 01:00.740 --> 01:04.400 There is one other thing I want to mention, which is the Bernoulli distribution. 01:04.610 --> 01:10.370 So we have the uniform distribution, which will give a range of integers or reals, which are all equally 01:10.370 --> 01:14.180 probable. With the Bernoulli distribution, we get Booleans. 01:14.900 --> 01:17.870 So all the random numbers are scaled to give 01:17.870 --> 01:19.970 either 1 or 0. True or false. 01:21.680 --> 01:22.860 And this works the same way. 01:22.880 --> 01:29.510 So we have a Bernoulli distribution object, and we pass the random number engine to it, and that will return 01:29.510 --> 01:31.790 a random number, or a random Boolean. 01:32.780 --> 01:38.510 So this is very useful if you want to make decisions which are 50-50. You know, equivalent to tossing 01:38.510 --> 01:38.930 a coin. 01:41.270 --> 01:43.910 Okay, so let's see how things are going in my kingdom, today. 01:45.680 --> 01:50.060 "Your subjects are grateful for your wise and benevolent rule". And so they should be! 01:53.410 --> 01:57.910 As this is rather a short video, I thought it was worth having a look at how the shuffle algorithm is 01:57.910 --> 01:58.510 implemented. 01:59.620 --> 02:04.720 The basic idea is that you take an elements and then pick another element at random and then swap the 02:04.720 --> 02:05.080 two. 02:05.680 --> 02:10.000 So you have a loop where you go through, and you swap each element with another one, which is chosen at 02:10.000 --> 02:10.450 random. 02:14.350 --> 02:16.090 So here is this code. 02:16.540 --> 02:18.130 Here is my attempt at implementing it. 02:20.330 --> 02:27.410 So we have our loop in which we swap the elements, so we iterate over each element and we choose 02:27.410 --> 02:28.520 another element at random. 02:28.970 --> 02:34.010 So we are going to use a random number as the index of the element that we are swapping it with. 02:35.780 --> 02:38.560 We use the uniform_int_distribution, with ints. 02:40.190 --> 02:44.180 This is going to generate an index, so we need to make sure it is within range. 02:44.630 --> 02:47.150 We do not want any out of bounds access errors. 02:48.320 --> 02:53.750 So the range we need is from 0 to size() minus one, the number of elements minus one. 02:54.410 --> 02:57.020 And then we just loop over all the elements and swap them. 03:00.010 --> 03:01.180 Okay, so there we are. 03:01.760 --> 03:07.530 Our vector is shuffeld. The real implementation is going to be a templated function, of course, 03:07.930 --> 03:13.720 and it will have some code to decide whether to use the real or integer version of uniform distribution. 03:14.530 --> 03:19.540 The mathematics might be a bit more sophisticated. But, anyway, this will give you an idea of what 03:19.540 --> 03:19.870 it does. 03:20.980 --> 03:22.600 Okay, so that is it for this video. 03:23.080 --> 03:23.950 I will see you next time. 03:24.160 --> 03:26.020 Until then, keep coding!