1
00:00:00,510 --> 00:00:04,790
‫Now, we have prepared our data for image classifier.

2
00:00:06,390 --> 00:00:09,330
‫The next step is model creation.

3
00:00:10,350 --> 00:00:14,700
‫There are two ways to create and train your model in keras.

4
00:00:15,390 --> 00:00:21,510
‫The first one is the sequential model api and the second one is the functional api.

5
00:00:24,290 --> 00:00:32,790
‫Sequential api is a straight forward and simple, whereas functional api is little bit complex, but

6
00:00:32,790 --> 00:00:40,110
‫it will give you the flexibility to create some complex neural networks. Sequential

7
00:00:40,140 --> 00:00:48,900
‫Api is useful to create layer by layer models such as these, where all the outputs of previous layer

8
00:00:49,050 --> 00:00:53,040
‫are connected as inputs of the next layer and so on.

9
00:00:53,760 --> 00:00:59,420
‫So for a simple neural network like this, sequential api is recommended.

10
00:01:02,920 --> 00:01:12,000
‫But for some advanced complex structure such as this, here we are using input as an input for concat

11
00:01:12,010 --> 00:01:12,790
‫layer as well.

12
00:01:13,600 --> 00:01:16,000
‫So this layer have two inputs.

13
00:01:17,140 --> 00:01:23,110
‫One, all of our primary input parameters and the outputs of hidden layer as well.

14
00:01:24,340 --> 00:01:33,370
‫So anywhere if you want complex structure like this or you want to only use some part of your input

15
00:01:33,460 --> 00:01:40,120
‫in one hidden layer and other part of input in some other hidden layer, for all such variations, you

16
00:01:40,120 --> 00:01:41,500
‫can use functional api.

17
00:01:42,340 --> 00:01:48,160
‫But for a straight forward, dense neural networks like this, you can use sequential api.

18
00:01:50,000 --> 00:01:53,180
‫In this course first, we will use the sequential api.

19
00:01:53,810 --> 00:01:57,640
‫Then we will also look at this example of functional api.