ARTIFICIAL INTELLIGANCE
|
|
|
7 points |
(2 points) (due: 5 pm, 9/28) This question is not a programming question. Compute to change the given connection weights by backpropagation algorithm. There are two input units, three hidden units, and one output unit. The input training data is 1 and 1. The target output is 1.
The learning rate is 0.5. The weights are given as between input unit 1 and hidden unit 1 is 0.1 between input unit 1 and hidden unit 2 is 0.2 between input unit 1 and hidden unit 3 is 0.3 between input unit 2 and hidden unit 1 is 0.4 between input unit 2 and hidden unit 2 is 0.5 between input unit 2 and hidden unit 3 is 0.6 between hidden unit 1 and output unit 1 is 0.7 between hidden unit 2 and output unit 1 is 0.8 between hidden unit 3 and output unit 1 is 0.9
Type your answer and submit it
(5 points) (due: 5 pm, 10/02) This question is a programming question. You can use your favorite computer language. Do not use Machine learning libraries for this question.
Program a multi-layered ANN to simulate the following Input and Output operation. You can use backpropagation algorithm to update the connection weights of your network.
|
Input attribute 1 |
Input attribute2 |
Target output |
|
0 |
0 |
1 |
|
0 |
1 |
0 |
|
1 |
0 |
0 |
|
1 |
1 |
1 |
When your actual output of output unit is larger than 0.9, you can accept it as 1. When your actual output of output unit is less than 0.1, you can accept it as 0.
You need to decide your learning coefficient. You will start with your small random initial weights for this problem. After your training process is done, you need to print your final actual output (produced from Sigmoid function) for each input to verify your result is close enough to target output.
Test your system with some different architecture.
First, simulate it with 2 input units, 2 hidden units, and 1 output unit.
Next, simulate it with 2 input units, 3 hidden units, and 1 output unit.
Next, simulate it with 2 input units, 3 hidden units, and 2 output units.
You need to submit your source file and a printed output from your program