DOC PREVIEW
UW-Madison ECE 539 - Using a MLP to Determine Whether a Signature is Forged or Valid

This preview shows page 1-2 out of 7 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 7 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 7 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 7 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Using a MLP to Determine Whether a Signature is Forged or ValidBy Kevin PontoFinal Project for ECE 539At the University of Wisconsin- MadisonWhat are the real world implications of this project?Signatures are used to verify a person is who they say they are.Things like credit cards, checks use signatures to verify idThings like contracts use a signature to the contract legally bindingFinding forges could save banks and credit card companies lots of moneyGathering the DataSign my name into a computer 100 times in a 2x6 inch imageHave 2 other people sign their name once, sign my name 4 times and then try to forge my signature while looking at my signature as a guideFirst:Then:realrealforgedDo Dimension Reduction on the InputsThis makes it so we have a reasonable amount of inputsRun the Inputs on an MLPFound optimal parameters areAlpha = .01Mom = .8Epochs = 1000Epoch size = 64Checks every 10 EpochsWith these parameters the MLP converges in < 300 Epochs !The MLP has lots of problems with this forged signatureIf the data is trained with one persons forges and tested with the other persons forges worst case scenario it is classifies 22% of the samples incorrectlyResultsValid SignatureForged SignatureConclusionThe MLP preformed quite well compared to both of the baseline models. If we look at real world scenarios, having absolutely no data on the forger, then the model was accurate 80-90% of the time. What makes these results useful is that the MLP was 100% accurate when claiming a signature was forged. This means that if someone were to implement this style in a bank to check for forged bank checks, if it ever flagged a signature as being forged, it would be a good idea to double check to see if is a valid transaction since the MLP appears not to give false positives in this scenario. The toughest part of this would be generating good forged signatures to test over. Other then having a person forge a signature several times for input, the MLP will not work. The good news is that if one can get this data, the MLP seems to work to detect forges signed by someone


View Full Document

UW-Madison ECE 539 - Using a MLP to Determine Whether a Signature is Forged or Valid

Documents in this Course
Load more
Download Using a MLP to Determine Whether a Signature is Forged or Valid
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Using a MLP to Determine Whether a Signature is Forged or Valid and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Using a MLP to Determine Whether a Signature is Forged or Valid 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?