1
00:00:01,020 --> 00:00:02,490
Welcome back.

2
00:00:02,490 --> 00:00:08,680
Earlier we mentioned there are generally three activities involved in training a machine learning model

3
00:00:09,090 --> 00:00:12,120
predict compare and learn.

4
00:00:12,330 --> 00:00:17,900
In the previous lessons we saw how prediction is done and its lesson which will talk a bit to bowl the

5
00:00:17,910 --> 00:00:18,810
comparison.

6
00:00:20,760 --> 00:00:25,510
Let's take a simple single neural network again.

7
00:00:25,530 --> 00:00:26,320
We learned.

8
00:00:26,640 --> 00:00:30,000
We predict by multiplying the input by the weight.

9
00:00:30,000 --> 00:00:34,760
Now we are going to introduce another parameter known as error.

10
00:00:34,950 --> 00:00:42,650
We find the error by subtracting the value we expected from the value we've predicted and then squaring

11
00:00:42,690 --> 00:00:48,790
the answer we require because we want the answer to always be positive.

12
00:00:49,260 --> 00:00:58,710
Yes an example let's say the weight is zero point eight on the expected value from what empirical measurement

13
00:00:58,800 --> 00:00:59,900
is twenty six.

14
00:01:00,120 --> 00:01:02,940
And our inputs is twenty five.

15
00:01:02,940 --> 00:01:13,990
If we perform our prediction of course input times weight we get 25 multiplied by zero point eight and

16
00:01:14,230 --> 00:01:22,630
when we perform this multiplication we end up with 20 the predicted value is 20 and then we have 26

17
00:01:22,660 --> 00:01:25,350
as the expected value to compute the error.

18
00:01:25,360 --> 00:01:34,570
We simply do 20 minus 26 and we can uh we squared a difference 20 minus 26 gives us minus six minus

19
00:01:34,570 --> 00:01:35,980
X squared is 36.

20
00:01:36,460 --> 00:01:40,810
So this arrow shows that it's not correct.

21
00:01:40,810 --> 00:01:45,700
So we've got to re adjust the weight in order to minimize the error.

22
00:01:46,120 --> 00:01:49,710
And we shall see how all of this is done programmatically.

23
00:01:49,780 --> 00:01:51,520
So this all there is for this lesson.

24
00:01:51,520 --> 00:01:52,620
I'll see you in the next lesson.