1 00:00:01,290 --> 00:00:04,030 ‫Let's start with perceptions. 2 00:00:05,070 --> 00:00:06,410 ‫Like in biology. 3 00:00:06,720 --> 00:00:13,370 ‫A single cell of our nervous system is called a neuron in artificial neural networks. 4 00:00:13,500 --> 00:00:23,070 ‫One of the earliest such artificial neuron was a perception perception was the lab in 1950s. 5 00:00:23,070 --> 00:00:28,900 ‫Yes the work on neural networks began nearly 70 years ago. 6 00:00:29,100 --> 00:00:35,940 ‫Today we use other models of artificial neurons such as sigmoid neurons but to understand take my neurons 7 00:00:36,210 --> 00:00:38,890 ‫we need to first look at perception. 8 00:00:40,360 --> 00:00:45,150 ‫Here's a simple pictorial representation of how perception works. 9 00:00:45,660 --> 00:00:56,100 ‫Perception is this circle or a black box which takes in several binary inputs x1 x2 x3 and so on. 10 00:00:56,100 --> 00:01:07,530 ‫DL exam and produces a single binary output represented by byte by binary input and binary output. 11 00:01:07,620 --> 00:01:19,160 ‫I mean that these variables can only take two values for example zero and one true or false etc.. 12 00:01:19,170 --> 00:01:29,700 ‫There are several ways in which these x1 x2 x3 can give us the desired output by one of the rule is 13 00:01:32,400 --> 00:01:44,460 ‫that we will multiply each of these input values with weight W1 W2 W3 and then compare if the final 14 00:01:44,460 --> 00:01:52,320 ‫value of the sum of these products is greater than a threshold value or not. 15 00:01:52,770 --> 00:01:59,710 ‫If the sum value is greater than the Perseid drone gives an output value of 1. 16 00:02:00,120 --> 00:02:07,650 ‫And if it is less than say sure it gives out an output value of 0. 17 00:02:07,650 --> 00:02:11,790 ‫Mathematically this is how we represent this logic. 18 00:02:11,790 --> 00:02:16,560 ‫This is the summation of weight with feature values. 19 00:02:16,590 --> 00:02:30,720 ‫Basically this means X1 input w 1 plus X2 into W2 plus x3 into W3 and so on the exam into W M the sum 20 00:02:30,720 --> 00:02:34,970 ‫of all these products is this left random. 21 00:02:35,130 --> 00:02:38,540 ‫We compared this sum the threshold value. 22 00:02:38,700 --> 00:02:42,600 ‫If this is less than the threshold we give an output of zero. 23 00:02:43,440 --> 00:02:46,440 ‫If it is more than the threshold we give output of 1 24 00:02:51,070 --> 00:02:57,730 ‫Let's take a simple example which may not be very realistic but you will get the idea of how this perception 25 00:02:58,000 --> 00:03:00,900 ‫functions. 26 00:03:00,910 --> 00:03:08,470 ‫Let's say you want to decide whether you should parties is a particular shirt or not. 27 00:03:08,470 --> 00:03:17,710 ‫You might make your decision by weighing up three factors whether this shirt is blue or not whether 28 00:03:17,710 --> 00:03:27,580 ‫the shirt is falsely or half sleeved and whether the fabric is cotton or not we can represent these 29 00:03:27,580 --> 00:03:31,820 ‫three variables using three binary variables. 30 00:03:32,740 --> 00:03:36,580 ‫For instance X1 is equal to one. 31 00:03:36,670 --> 00:03:39,050 ‫If the shirt is blue and it is 0. 32 00:03:39,100 --> 00:03:43,430 ‫If it is not blue x2 is equal to 1. 33 00:03:43,450 --> 00:03:53,350 ‫If it is fully believed and 0 if it is half sleeve and extra is equal to 1 for cotton fabric and 0 for 34 00:03:53,470 --> 00:03:57,130 ‫non cotton fabric. 35 00:03:57,160 --> 00:04:04,720 ‫Now suppose that you absolutely adored blue collared shirt and you would prefer food sleeved cotton 36 00:04:04,720 --> 00:04:11,530 ‫fabric shirt much Cleveland and fabric is not as important as the color of the shirt. 37 00:04:13,570 --> 00:04:20,840 ‫So here are a sample rate of importance that you assign to these features. 38 00:04:20,890 --> 00:04:25,210 ‫You give weight of 7 to the shirt color. 39 00:04:25,800 --> 00:04:30,690 ‫I have replaced the value of w 1 with this number 7. 40 00:04:30,910 --> 00:04:36,550 ‫We assign a weight of 4 to Cleveland and a weight of 2 to the fabric. 41 00:04:38,110 --> 00:04:47,140 ‫Finally we also take a threshold value of 8 to decide whether two parties pressured or not. 42 00:04:47,140 --> 00:04:49,740 ‫With these choices of weights and threshold. 43 00:04:50,020 --> 00:04:53,910 ‫Let's see which of these three showed would we buy. 44 00:04:56,320 --> 00:05:04,010 ‫So far this first shirt we have blue in the first column which signifies the color of the shirt. 45 00:05:04,120 --> 00:05:08,880 ‫It is half sleeved so half and the second column it is non garden so non Gordon. 46 00:05:08,880 --> 00:05:16,570 ‫The third column which is what fabric the fourth column is for calculation of some. 47 00:05:17,050 --> 00:05:22,440 ‫As I told you previously we calculate the product of weights the features. 48 00:05:23,080 --> 00:05:29,170 ‫Add them together to find we calculated some in the Fifth Column. 49 00:05:29,170 --> 00:05:35,020 ‫We have written the threshold value that is pre decided in the sixth column. 50 00:05:35,020 --> 00:05:38,560 ‫We compare this some value with detail shown value. 51 00:05:38,950 --> 00:05:42,420 ‫If the sum is greater than threshold we will buy a t shirt. 52 00:05:42,430 --> 00:05:45,630 ‫If the sum is less than threshold we will not buy a t shirt. 53 00:05:47,140 --> 00:05:49,430 ‫So let let's see what happens with this. 54 00:05:49,500 --> 00:05:50,800 ‫Shirt. 55 00:05:51,000 --> 00:05:54,630 ‫The fresh shirt is blue in color but blue. 56 00:05:54,730 --> 00:05:57,970 ‫We have X1 value of 1 for not blue. 57 00:05:57,970 --> 00:06:03,760 ‫It would have been 0 so X1 is 1 x2. 58 00:06:03,880 --> 00:06:08,370 ‫We just leave is zero because it is half sleep. 59 00:06:08,490 --> 00:06:13,300 ‫Fabric is non cotton which is again zero. 60 00:06:13,650 --> 00:06:15,610 ‫We find out the same product. 61 00:06:15,710 --> 00:06:27,310 ‫7 It is rated for color multiplied by the value of X 1 which is one plus 4 wear where for sleeves multiplied 62 00:06:27,310 --> 00:06:35,770 ‫by the value of x2 which is zero because it is half sleep plus 2 which is weighted for fabric multiplied 63 00:06:35,770 --> 00:06:40,720 ‫by the value of fabric which is zero because it is not cotton. 64 00:06:41,050 --> 00:06:44,040 ‫The final time we get is 7. 65 00:06:44,110 --> 00:06:48,870 ‫We compare this some value with the threshold value which is 8. 66 00:06:49,000 --> 00:06:50,650 ‫The sum is less than 8. 67 00:06:51,490 --> 00:06:54,560 ‫So we are not going to buy this shirt. 68 00:06:56,220 --> 00:06:58,970 ‫Let us do this activity for the second shirt. 69 00:06:59,080 --> 00:07:03,850 ‫The second shirt is blue colored full sleeved non Gordon fabric. 70 00:07:05,020 --> 00:07:13,190 ‫If you repeat the calculation the only difference is going to be the value of x2 for the full sleeve 71 00:07:13,190 --> 00:07:13,590 ‫shirt. 72 00:07:13,630 --> 00:07:20,200 ‫Extra will be 1 so 7 plus 4 is going to come out as eleven eleven is more than eight. 73 00:07:20,200 --> 00:07:23,920 ‫So we are going to buy this shirt. 74 00:07:24,130 --> 00:07:29,030 ‫Similarly for the third shirt which is not blue full sleeved cotton shirt. 75 00:07:29,170 --> 00:07:35,740 ‫Some comes out to be 6 which is less than 8 which means that we are not going to buy this shirt. 76 00:07:35,830 --> 00:07:44,440 ‫Can you now see how perception is deciding the output that whether you will buy a shirt or not. 77 00:07:44,440 --> 00:07:50,370 ‫It is just multiplying the values of the feature with corresponding weight and checking some against 78 00:07:50,370 --> 00:07:54,730 ‫the threshold value if the sum is larger than the threshold. 79 00:07:54,980 --> 00:08:00,860 ‫It gives one output if it is smaller then it gives other output. 80 00:08:00,920 --> 00:08:07,300 ‫This is a very simple example which I have given to make you understand how a perception is working. 81 00:08:07,400 --> 00:08:15,500 ‫In reality we solve much more complex problems in which we have numerous input variables and many conditions. 82 00:08:15,500 --> 00:08:19,670 ‫We will get to them in the due course of the lectures. 83 00:08:19,940 --> 00:08:29,540 ‫As you can see a perception requires these weights and this threshold value to give out an output. 84 00:08:29,540 --> 00:08:33,750 ‫And how will perception get the values of these parameters. 85 00:08:33,920 --> 00:08:38,710 ‫One ways we give the values in which case it is not learning. 86 00:08:38,960 --> 00:08:42,280 ‫It is simple programming. 87 00:08:42,290 --> 00:08:50,510 ‫The other ways learning when we provide deeper separation with historical data of which shots were selected 88 00:08:50,990 --> 00:08:57,830 ‫and which shirts were rejected and the perception decides debate and traditional value according to 89 00:08:57,830 --> 00:09:00,330 ‫that previous data. 90 00:09:00,680 --> 00:09:07,400 ‫By the way you can get different models by changing weights and to be sure. 91 00:09:07,610 --> 00:09:19,830 ‫For example if you want to select a shirt which is blue food sleeved and cotton only and no other combination. 92 00:09:19,820 --> 00:09:28,400 ‫This set of weight and threshold ensures that right out you will see only blue cotton full sleeve shirt 93 00:09:28,640 --> 00:09:35,750 ‫will be selected and no other shirt will pass through. 94 00:09:35,750 --> 00:09:38,540 ‫So that's about it. 95 00:09:38,540 --> 00:09:45,260 ‫This is a basic introduction to the perception we will extend this idea of perception in the next lecture.