WEBVTT 00:00.080 --> 00:01.080 Welcome back. 00:01.120 --> 00:03.720 We learned about fit function. 00:03.760 --> 00:10.320 We passed the four parameters features, targets, epochs, and verbose. 00:10.640 --> 00:14.680 Now we need to get the trained weights and bias. 00:14.800 --> 00:25.280 At first we need to run all in order to train well this model and wait for the cell to be executed. 00:25.320 --> 00:26.200 And here we go. 00:26.320 --> 00:28.840 The cell is executed correctly. 00:28.880 --> 00:36.920 Now there is a very important thing we need to do which is evaluating the trained weights and bias. 00:36.920 --> 00:47.480 So let's see if the, uh, that the trained weights and bias are close to the equation that we put with 00:47.520 --> 00:50.400 the corresponding data frames. 00:50.400 --> 00:59.800 If we remember that price equals to 2.9 times size plus 2.17, this is from the previous example with 00:59.800 --> 01:05.360 uh, and by the way, this example, the TensorFlow example uses the same data. 01:05.360 --> 01:08.430 So we have the slope is 2.9. 01:08.470 --> 01:12.550 The bias is and the intercept is 2.7. 01:12.870 --> 01:13.390 Okay. 01:13.550 --> 01:20.750 Now let's test those and display the result of bias and the weight from here. 01:20.790 --> 01:28.910 Remember we have the equation y equal to w x plus b y is w and bias is b. 01:28.910 --> 01:34.070 So let me create two variables to be displayed weights and bias. 01:34.190 --> 01:38.590 And thanks for the autogenerated code by the Colab. 01:38.670 --> 01:39.390 Bias. 01:39.550 --> 01:45.630 Weights and bias are two variables assigned by the model layers. 01:45.790 --> 01:47.390 Dot get weights. 01:47.390 --> 01:57.310 Here we are displaying the trained equation and learned weights and bias model dot layers zero index 01:57.350 --> 02:02.030 accesses the first and only layer of our model. 02:02.030 --> 02:07.190 Since our model contains only one layer, we need to access only this first layer. 02:07.350 --> 02:08.710 Get weights. 02:08.750 --> 02:13.390 Returns the learned parameters as the list. 02:13.390 --> 02:15.230 So we have a list here. 02:15.710 --> 02:17.350 Weights of and bias. 02:17.390 --> 02:25.190 We need to access them and display them again guys, in order to get the learned weights and bias, 02:25.230 --> 02:27.870 we use model dot layers. 02:27.870 --> 02:36.190 We are accessing the first layer and only one layer we have here in this simple simple trained model 02:36.350 --> 02:38.390 and get weights. 02:38.390 --> 02:41.470 Function returns the learned parameters as a list. 02:41.470 --> 02:47.670 Weights and bias w equal or y equal w times x plus b. 02:47.750 --> 02:53.470 This is the general formula for linear regression and linear relationships w the weights. 02:53.470 --> 02:57.870 The slope parameter, which is a 2D array shape. 02:57.870 --> 03:06.910 So for that, in order to access it and print it, we use weights and access zero and then zero. 03:06.950 --> 03:08.630 This is a 2D array. 03:08.670 --> 03:17.980 I am accessing this 2D array and the bias get the intercept parameter which is 1D array shape. 03:17.980 --> 03:23.860 So since the bias is 1D array, we use this and access the first index. 03:23.860 --> 03:29.740 And since the weights is a 2D array, we access it like this for. 03:29.780 --> 03:33.860 Let's run and print those parameters and here we go. 03:34.140 --> 03:40.300 The weight equals to 3.005 which is close to 2.9. 03:40.460 --> 03:46.460 And the bias intercept is 2.04 which is close to 2.7. 03:46.660 --> 03:48.580 So congratulations guys. 03:48.780 --> 03:53.260 We trained the model and we get those parameters. 03:53.300 --> 03:57.500 Now let's use those parameters to make the predictions. 03:57.500 --> 04:01.380 So here let me show you guys make predictions. 04:01.380 --> 04:05.340 Predictions equals to model dot predict. 04:05.620 --> 04:08.740 And here pass the x value. 04:08.860 --> 04:13.700 Compare with the original relationship three times size plus two. 04:13.980 --> 04:17.580 Here I am printing the prediction. 04:17.860 --> 04:26.180 The learned relationship price equals to weight times size plus, let me see the original relationship. 04:26.220 --> 04:29.380 Equal price equals to three times size plus two. 04:29.420 --> 04:32.700 This is the, uh, the original relationship. 04:32.940 --> 04:46.460 The learned relationship after we provide the data is 3.0015 times size and plus two, which is 2.0467. 04:46.460 --> 04:46.940 Okay. 04:46.980 --> 04:48.740 So congratulations guys. 04:48.940 --> 04:56.060 We trained our model in a very close manner to get the original relationship, which is three times 04:56.100 --> 04:57.180 size plus two. 04:57.340 --> 05:05.500 In the previous example, we get a close to three which is 2.9 and 2.7. 05:05.660 --> 05:10.540 Also this uh TensorFlow model is well trained. 05:10.740 --> 05:14.620 And congratulations we get this amazing result. 05:14.780 --> 05:21.620 And we get this prediction with original close to the original relationship.