Closed

# Matlab coding needed along with report

This project was awarded to bchandra1955 for \$52 USD.

\$30 - \$250 USD
2
###### Project Description

Need this to be done..Let me know if u can do do this in the best way

Start, I am attaching the midterm and sample data given by teacher. 64 columns, 8 X 8 matrix. I will require the code and results in the word document.

Some explanation

Each row in the text file in our input. Lets assume that you train the network to identify Zero. So reading the first line from the [url removed, login to view] file you will get the 64 input values which are the Xi's. As you are reading from the [url removed, login to view] you know what the ground truth is which is simply zero. That's the 'Oi' in the algorithm. Run the Backpropagation function now. I don't know what the Matlab function takes in as input.

As we have 10 output nodes (node1 for 0, node2 for 1..), for this particular example with zeros your neural net will be trained to output 1 from node1 and 0 from all the other 9 output nodes.

After doing this for all 51 rows you will be having the final weights associated with each nodes. Now test you model using the [url removed, login to view] and calculate the misclassification (by misclassification I mean 1 in any other nodes other than node1).

Repeat the same steps to train other digits.

Hey,

yes, it should tell us # of time or percentage that our model caclulated or identified the right #. If we input [url removed, login to view] then the 1st node should be 1 (turned on ) and rest zero for all other 9 output nodes. I hope this helps

I think digits 0 to 9 only. Yes, train the system and calculate the error. We have to use backpropagation algorithm.

some more clarification for your help

Once you build the model on the training data we need to test the model on the test data and see how it classifies the data. The number of misclassification on this test data gives you the error of the model on the test data.

For the final report I think we need to show him the error on the training example(norm of error), our stopping criteria which is based on the error. I'm simply going to set the threshold for the stopping criteria before starting the algorithms. In order to compare the model with 2,3 and 4 hidden units (question 4) we need to find the error of the model on the test data

bid for less..

i can pay less than 50\$

i need the code asap and report after that...

## Hire Freelancers who also bid on this project

• Forbes
• The New York Times
• Time
• Wall Street Journal
• Times Online