1
00:00:11,080 --> 00:00:16,720
So in this lecture, we will summarize everything we learned in this section, as you saw, the section

2
00:00:16,720 --> 00:00:18,610
was mostly about two things.

3
00:00:19,120 --> 00:00:22,480
Number one, introducing you to TensorFlow syntax.

4
00:00:22,990 --> 00:00:28,780
So now you know all the main steps that go into writing a TensorFlow model, from defining a model to

5
00:00:28,780 --> 00:00:30,730
training it to making predictions.

6
00:00:31,360 --> 00:00:36,700
You learn that training with neural networks isn't as simple as calling model Dutch fit as it is with,

7
00:00:36,700 --> 00:00:42,940
so I can learn things to care as the process is greatly simplified, but it's still less automatic than

8
00:00:42,940 --> 00:00:44,470
most Saikia learn models.

9
00:00:45,550 --> 00:00:51,430
And number two, you saw that the linear models we learned about before are actually models of the neuron,

10
00:00:51,430 --> 00:00:55,090
which are inspired by biology as in real living creatures.

11
00:00:55,810 --> 00:01:00,760
This idea of the neuron is important because this was the inspiration behind deep learning.

12
00:01:01,480 --> 00:01:07,300
Unlike models such as the SVM and decision, trees which are purely mathematical neural networks are

13
00:01:07,300 --> 00:01:11,470
inspired by how biological brains work and how we think as humans.

14
00:01:12,040 --> 00:01:17,380
This is one aspect of deep learning that differentiates it from other forms of artificial intelligence.

15
00:01:21,960 --> 00:01:27,840
You also saw that, unlike psyche, learn understanding the learning process a bit more in depth is

16
00:01:27,840 --> 00:01:29,730
required when using deep learning.

17
00:01:30,390 --> 00:01:34,620
This is really important because as the models we learn about become more complex.

18
00:01:35,010 --> 00:01:38,300
Choosing the right hyper parameters tends to become more difficult.

19
00:01:38,850 --> 00:01:44,640
With linear models, it's easy to choose a learning rate, but with CNN's CNN's and Aunt ends, there

20
00:01:44,640 --> 00:01:46,320
are many more parameters to consider.