You are on page 1of 5

Written Assignment 7 Page 1 of 5

Written Assignment 7

University of the People

CS 4407: Data Mining and Machine Learning

Unit 7: Artificial Neural Networks – Part 2

<Name Removed>

This study source was downloaded by 100000813700581 from CourseHero.com on 12-29-2021 09:07:24 GMT -06:00

https://www.coursehero.com/file/56876675/CS-4407-Written-Assignment-7pdf/
Written Assignment 7 Page 2 of 5

For Unit 7, you will report the results from neural network training you completed in Unit 6. Write a

short paper explaining the process of developing the network provided including:

• Details on how many iterations of network designs were evaluated.

• What results were obtained?

• What alternatives were tested to determine the best approach for training a network that

would yield accurate results in the minimum of training steps?

When training the artificial neural network, it would be best to use a higher number of learning steps

compared to a low number of learning steps. With a higher number of learning steps, the machine

learning algorithm/function would be able to start minimizing the amount of error in determining the

end result. If we take a look at the error progress graph, we can see that with less number of training

steps, there is significantly more space between each of the data points, which show that there is

significant margin of error between each data point. As the machine takes more steps to learn, there is

less and less space and margin of error when predicting the end result, as shown by the error progress

graph as we move towards the right as the number of training steps increases.

In terms of iterations of network designs evaluated, we must first start with an arbitrary number of

training steps and can begin using the default learning rate, momentum, and weight range. Let’s pick the

learning rate of 0.3, momentum of 0.9, and weight range from -1 to 1. If we were to use 1000 training

steps, there is still a significant amount of space, and thus margin of error, between the data points. So,

to determine the minimum number of training steps needed, we must arbitrarily increase the training

steps until the error progress shows the data points barely starting to touch each other in the error

progress graph. In our case, if we are to set the number of training steps to 2500, we can start to see a

small portion of data point starting to cluster very closely together closer to the bottom of the graph.

This study source was downloaded by 100000813700581 from CourseHero.com on 12-29-2021 09:07:24 GMT -06:00

https://www.coursehero.com/file/56876675/CS-4407-Written-Assignment-7pdf/
Written Assignment 7 Page 3 of 5

We can also arbitrarily set the number of training steps to a little higher such as 5000, in which case, we

can use the error progress graph’s tick marks to mark off where the data points are starting to cluster

more together as the number of training steps increase. Setting the training steps to 5000 is a good

number as we can still see the large error margin in the left portion without having the clustered data

points extend too far on the right, as we cannot see the full graph if we were to use a large training

steps number like 10,000.

Figure 1: Example of error progress graph at 1000 training steps

Figure 2: Example of error progress graph at 10,000 training steps

This study source was downloaded by 100000813700581 from CourseHero.com on 12-29-2021 09:07:24 GMT -06:00

https://www.coursehero.com/file/56876675/CS-4407-Written-Assignment-7pdf/
Written Assignment 7 Page 4 of 5

Figure 3: Example of error progress graph at 5000 training steps

We must also consider what we would accept as the threshold in which there is acceptable error

between estimated data. In many cases, the acceptable threshold of error is within 5% or less of the

expected result, and taking a look at the results in the error progress graph, we can assume that the

minimum number of training steps needed should be somewhere between 2500 to 3000 steps. Using

5000 training steps as opposed to 10,000 or arbitrarily guessing a number, it will show similar results

and a better idea of the whole picture, which also produces the minimum number of training steps at

approximately between 2500 and 3000. Thus, in my case, I have gone through 3 iterations of network

design evaluations to determine the results.

This study source was downloaded by 100000813700581 from CourseHero.com on 12-29-2021 09:07:24 GMT -06:00

https://www.coursehero.com/file/56876675/CS-4407-Written-Assignment-7pdf/
Written Assignment 7 Page 5 of 5

References

James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An Introduction to Statistical Learning with

Applications in R. Retrieved from http://faculty.marshall.usc.edu/gareth-

james/ISL/ISLR%20Seventh%20Printing.pdf

Venables, W. N., & Smith, D. M. (2012). An Introduction to R. Retrieved from https://cran.r-

project.org/doc/manuals/R-intro.pdf

This study source was downloaded by 100000813700581 from CourseHero.com on 12-29-2021 09:07:24 GMT -06:00

https://www.coursehero.com/file/56876675/CS-4407-Written-Assignment-7pdf/
Powered by TCPDF (www.tcpdf.org)

You might also like