1
00:00:11,070 --> 00:00:16,590
So in this lecture, we'll be introducing the next section of this course, which is all about the artificial

2
00:00:16,590 --> 00:00:19,290
neuron and how to implement one in TensorFlow.

3
00:00:20,070 --> 00:00:25,260
The main goal of this section is to learn the basics of TensorFlow so that you can carry this knowledge

4
00:00:25,260 --> 00:00:28,860
forward in the later sections on and CNN's intranet.

5
00:00:29,550 --> 00:00:35,520
Basically, what we'll end up doing in this section is revisiting an old topic, which is logistic regression.

6
00:00:35,760 --> 00:00:41,970
But from a new perspective, this new perspective is that this model is actually a model of the neuron,

7
00:00:42,210 --> 00:00:46,110
and obviously a neuron is the basic building block of neural networks.

8
00:00:46,770 --> 00:00:52,080
In addition, we previously didn't discuss how logistic regression actually learns from the dataset

9
00:00:52,080 --> 00:00:52,710
we give it.

10
00:00:53,250 --> 00:00:59,190
So this section will look a little deeper into not just how the model makes predictions, but also how

11
00:00:59,190 --> 00:01:01,830
we train the model before we make those predictions.

12
00:01:06,310 --> 00:01:11,710
To give you a quick outline for this section, we'll begin by looking at the TensorFlow syntax for simply

13
00:01:11,710 --> 00:01:12,490
fitting a line.

14
00:01:13,150 --> 00:01:18,550
I'll assume that you've all encountered the concept of a line of best fit in high school so that this

15
00:01:18,550 --> 00:01:19,690
needs no introduction.

16
00:01:20,350 --> 00:01:23,860
This will introduce us to the very basics of TensorFlow syntax.

17
00:01:24,910 --> 00:01:30,730
The next step will be to look at how we can use the same syntax to implement logistic regression, which

18
00:01:30,730 --> 00:01:35,800
is another linear model, but for predicting categories instead of continuous values.

19
00:01:36,610 --> 00:01:41,890
Once we've done that, we'll go a little more in-depth into why this model is a model of the neuron

20
00:01:42,160 --> 00:01:43,870
and how this neuron learns.

21
00:01:44,440 --> 00:01:49,510
In other words, we'll talk a little bit about what goes on when we call model does fit instead of treating

22
00:01:49,510 --> 00:01:52,540
it like a magic box that does what we need it to do.

23
00:01:53,260 --> 00:01:58,840
This is important actually, because in deep learning, sometimes it does not do what we needed to do.

24
00:01:59,410 --> 00:02:04,900
This is unlike models in Typekit learn, which work pretty much most of the time in deep learning.

25
00:02:04,900 --> 00:02:09,850
It's necessary to monitor your model's progress to make sure things are going as expected.

26
00:02:10,449 --> 00:02:15,100
In other words, you cannot just trust that just calling model does fit will do what you want.