New User Special Price Expires in

Let's log you in.

Sign in with Facebook


Don't have a StudySoup account? Create one here!


Create a StudySoup account

Be part of our community, it's free to join!

Sign up with Facebook


Create your account
By creating an account you agree to StudySoup's terms and conditions and privacy policy

Already have a StudySoup account? Login here

Computer Organization

by: Mireya Heidenreich

Computer Organization CS 3843

Mireya Heidenreich
GPA 3.55


Almost Ready


These notes were just uploaded, and will be ready to view shortly.

Purchase these notes here, or revisit this page.

Either way, we'll remind you when they're ready :)

Preview These Notes for FREE

Get a free preview of these Notes, just enter your email below.

Unlock Preview
Unlock Preview

Preview these materials now for free

Why put in your email? Get access to more of this material and other relevant free materials for your school

View Preview

About this Document

Class Notes
25 ?




Popular in Course

Popular in ComputerScienence

This 39 page Class Notes was uploaded by Mireya Heidenreich on Thursday October 29, 2015. The Class Notes belongs to CS 3843 at University of Texas at San Antonio taught by Staff in Fall. Since its upload, it has received 18 views. For similar materials see /class/231412/cs-3843-university-of-texas-at-san-antonio in ComputerScienence at University of Texas at San Antonio.

Similar to CS 3843 at UTSA

Popular in ComputerScienence


Reviews for Computer Organization


Report this Material


What is Karma?


Karma is the currency of StudySoup.

You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!

Date Created: 10/29/15
Information Fusion in Multimedia Information Retrieval An Overview Jie Yu Department of Computer Science University of Texas at San Antonio J Yu Data and Vision Seminar Outline Information Fusion IF Introduction Model and Applications IF in Multimedia Information Retrieval Multimodality and Multiclassifier Case Study Fusion in Biometrics Our related study Interactive Boosting Discussion and Conclusion J Yu Data and Vision Seminar What is Information Fusion Fusion A merging of diverse distinct or separate elements into a unified whole Merriam Webster dictionary Information fusion is an Information Process dealing with the association correlation and combination of data and information from single and multiple sensors or sources to achieve refined estimates of parameters characteristics events and behaviors for observed entities in an observed field of view JDL 1999 It is sometimes implemented as a Fully Automatic process or as a HumanAiding process for Analysis andor Decision Support J Yu Data and Vision Seminar Why is fusion necessary The fusion of redundant information from different sources can reduce overall uncertainty and thus increase the accuracy of the system Multiple sources providing redundant information can also increase the robustness of the system The fusion of complementary information provided by different sources results in an information gain due to the utilization of multiple sources of information vs a single source The fusion of information from multiple sensors may provide more timely information either because of the actual speed of operation of each sensor or because of processing parallelism that may possibly be achieved as part of the integration process J Yu Data and Vision Seminar How Information Fusion Works Multiple types of data gt Related to things of interest To improve estimates about those things These Basic Ideas are Transferable to Many Types of Problems Complete Information Fusion System Model JDL Wequot Leve1 Leve2 Leve3 Em ferss39rnrg Processing Processing Processing nu W i F Ln Status or Situation Bcnign 39Critical Data Ba Leve4 Processing Management HumanEngrg Dei enseSurveill ances stem Adap ve Processing Logic cg support IDecisionraiding Dalabase systems 3 en Acu39vc sensor control 39Visua39iza ondisplay monitoring 39Fusion process control S stems J W Data and msmn sammar Fusion System Applications Military applications Autonomous weaponry employing multiple sensors Broad area surveillance systems employing single weapons platforms eg ships airborne surveillance or distributed sensor networks Fire control systems employing multiple sensors for acquisition tracking and command guidance Intelligence collection systems Indications and warning IampW systems the mission of which is to assess threat and hostile intent Command and control nodes for military forces J Yu Data and Vision Seminar Example FusionBased Automatic Object Recognition FUSED COMPONENT MATCH SCORES SAR ONLY 34 pose collected Predicted EO ijL 88 pose Collected Predicted SAR Component EO Component Match Scores Match Scores Generate Hypothesis eg BRDM2 34 pose articulation x Match Scores Predict Measurements Low High gt Evaluate ComponentLevel Match Actual vs Predicted I Select Hypothesis with Best Match 3 Yul Data and Sm Seminar Fusion System Applications Industrial and commercial applications robotics machine intelligence remote sensing image processing medical systems Eg Video retrieval Hallimodal jutr3quot Text Aspect AIME Aw Motion Aspect Image Asp myquot 7 V m i 1 Lurkm 39 I 39 L R xx Me a f x quot2515 a a 1 1 1 I Scripl VMEU I Iml uu m AuL IIIu l39w39luliull maga jd Libr f titles GTE Mutatlazla 1min Ollj cl Imlrt 77 77 quot 1 Marla Index I 1min 39 39 7 a b Multiple Modality 1fillet ulleclinn Analysis quot i r 7 quot I 39oighted Fusion of Search Resullsi i l l l l l Film Runka Lia i E Multimedia Information Retrieval Simple model Matching Decision Module Feature Extraction I I Database J Yu Data and Vision Seminar Information Fusion in MIR Feature Extraction Module Multiple features gtvectors Concatenated vector Feature Fusion more discriminating hyperspace can be found in the new vector Matching Module One types of classifiers for multiple features or Multiple types of classifiers for one feature or Both The output score can be combined Decision Module The output decision of each classifier can be combined J Yu Data and Vision Seminar Two Forms of Information Fusion in MIR Multi Modality Modality refers to the type of communication channel used to convey or acquire information It also covers the way an idea is expressed or perceived or the manner an action is performed audio video text Eg Video clip gtvisual information audio information textural information Multi modality fusion occurs at feature extraction module Sin le source information may be represented by mu tiple features Eg Color image gtcolor texture shape J Yu Data and Vision Seminar Two Forms of Information Fusion in MIR MultiClassifiers Ensemble of Classifiers EOS A set of classifiers are trained to solve the same problem Applied on single or multiple source of information A single type of base classifiers or Different types of classifiers Bayesian KNN SVM Multiple classifiers can be generated by Randomize training set Randomize classifier structure I Or both Advantages High accuracy Can deal with extremely large data sets eg biology data Weightin schemes may provide an a proach for dealing with skewed ata 10001 business frau internet security Plenary Talk L 0 Hall ICPR 2006 Fusion Schemes The prediction of multiple classifiers need to be integrated into one fused decision by Fusion Scheme The output of different classifiers need to be normalized Rule based Decision is made by a simple operation on the output of all classifiers Eg Max Min Sum Mean Matching Module Eg AND OR Decision Module Learning based The out ut of all classifiers is fed into a learning process to obtain t e final decision Eg Decision Tree Neutral Network No one is uaranteed to be the best empirically or theoretica Iy J Yu Data and Vision Seminar Comparison of Simple Fusion gtTest conducted on 3 UCI benchmark data sets gtPCA and LDA Bayesian classifiers are fused gtSimple fusion scheme does not always works gtNo fusion scheme is guaranteed to be the best Schemes Probability Based PCA and LDA Fusion Tests Best Error Rate Thre sh01 d P Benchmark Fusion Method Heart BreastiCancer Banana 0184 02779 04272 PCA no fusion 06 06 06 0159 025 04425 LDA no fusion 06 07 06 07175 03065 04473 AND 01 01 02 quotORz 06135 02981 04432 01 02 04 Mean 0155 02688 04333 05 07 06 quotMaxz 01575 02682 04303 07 07 06 quotMinz 01735 02597 04194 05 06 06 J Yu Data and Vision Seminar Information Fusion in Biometrics What is Biometrics General definition the science and technology of measuring and analyzing biological data In information technology biometrics refers to technologies that measure and analyze unique human body characteristics eg Fingerprint Facial image Hand Geometry Handwriting Iris image Voice Information Fusion in Biometrics Two modes Enrollment mode Biometrical information is obtained and stored in database with identification label Authentication mode Identification Is heshe in the database Authentication Is heshe the one claimed Performance metrics Given a threshold False accept rate False reject rate A K Jain et al PRL 2003 Information Fusion in Biometric System F gar m Template Matching Module Decixiun Mudulll m 11mm FM quot E m m Mulching Module Denimquot M011qu il lammm Mallula Incc Fm Templaiex J W Data and ViSiOi v Seminar Experiment Settings Multi Modality Face Fingerprint Hand Geometry of 50 persons Independence among different biometrics is assumed Classifier Distance based Scores have to be normalized Performance measure ROC curve J Yu Data and Vision Seminar Example of Fusion in Biometrics Face Verification Challenges illumination expression head pose background Example of Fusion in Biometrics Fingerprint Verification aPreprocess to get ridge map bDetect ridge bifurcation and 1mm BIFURCATION ending cFind Minutiae point R RIDGEENDIHG dAign and find matching minutiae points between two finger prints J Yu Data and Vision Seminar Example of Fusion in Biometrics Hand Geometry Scores obtained from different biometric features Genuine Spares Hand Ga mew Emma Impcetar Ecore3 quotHR xx v quot inn 57351 HEELe EU Fingerprint Scaras R 39mquotquotFFP 4 Face Scares 9 s 5 III Performance of Three Biometrics I I IIIIIII I IIIIIIII I IIIIIIII I I IIIIIII I I I i 33quot 9 M yet 7quot j 3 Fingerprint tquotquot 1 r39 1 IE E3 f I H H In if aquot Iquot 355339 Face 3 a Ea quot Iquot Gaming Accept Hate 194 6x Hand Geumetw 33 I I lllll I I IIIIIII I I IIIIIII I I IIIIII I I IIIIII 1339 10 13quot 10quot 1 12 False MceptFlrJtei i Fusion of Face and Fingerprint 1C er quotW H r quotquot g5 Face Fingerrprint 4a quot 39 gt39 5 4 41 39 3PA 90 cr 4quot 63 wry m 85 farquot I E 4quot I U l 5 30 kegF Ix 3 I g 75 requot 39 m Fingerprint E m c ED 3 lt5 5539 Face i 1 ED 55 5r mI 1 F3 1c4 10quot m 1 0 10 False Accept Rate 23 J Yu Data and Vision Seminar Fusion of Face and Hand Geometry m3 a B E n 4r 4 Bquot r39 f gm Face Hand Geujmatry39 Er Effr43 z2 3 r z m Genume AmeptRaIe d Ln 1 21 I 39r1 g i I l 40 Ff Hand Geomatry 1U 1039 10 10 1a 11 False Accept Rate 963 J Yu Data and Vision Seminar Fusion of Fingerprint and Hand Geometry 1m 7 quoti a I thaw i ngerprint EHand GeomeIQquotquotn e QF quot quot 39 39 an quotquot 4 51 L V f rquot 4f 3quot 1 Bo V J k v quotxii 5quot 39 f E T Fingerpnni x g U gt t ECquot 391339 5 I g 50 39L G 3sz 539 39 Hand ileumng 30 39 mi 39J I A I 1 I U I V quot 10 ii m m 10 14 False Accepi Flame 2H J Yu Data and Vision Seminar Fusion of Three Biometrics 1m quot39 397quotECAquotF39 39 quot n 441nvni quot 39 L 39 Face Fingarprint 4 Hand Gaumairy39 gal3quot 90 quot39 3 quot aquot r 39r s ED a quot f it I 2 m quot Fi e rlnt 53 7390 quot 539 VP E Farm 3 3339 d 60 K II E 50 39 a a 40 39 Hand Geomairy ED 20 quotI IG 1 i quotI i T 339 10quot m m m m m Faise Accept Hale 2 J Yu Data and Vision Seminar Our Related Study Interactive Boos ng Motivation Multimodality and Multiclassifier system performs better in many MIR systems Compared with the singlesource singleclassifier system Fusion schemes selection is difficult and may not be efficient Some fusion scheme cannot improve the performance Human aid in a fusion process may provide more information gain in a guided direction Idea to introduce human centered computing in information fusion Yu Lu Xu and Tian ICASSP 2007 Strength of Human and Machine Machine Human cognitive Sc39tence statisttcal Analysis Perception Sltierm c Visualization Vtsuat mtemgence Data Mmmg V Information Vtsuahzatton Dasign Dem ston ScienCe ompresston amp Fittering User39s Vision Graphs and Rendenng Human Machine Interface Yu Lu Xu and nan xcnssv 2am Human Centered Computing and Information Fusion Human Centered Computing Combining the strength of human and machine Human strength creativity solving of ambiguity strategies and principles etc System strength accuracy without fatigue data storage deterministic processing Division of Labor between human and labor Enhancement of overall system performance Goal To amplify and extend of human cognitive and perceptual ability in information fusion Yu Lu Xu and Tian ICASSP 2007 Interactive Boosting Idea ofBoosUng Reinforcement training Weight combination Bachdea User relevance feedback on unlabeled data is incorporated into a iterative boosting process Semisupervised vs fullysupervised boosting Adaptive Discriminant Pro39ection ADP is used to generate multiple representation 0 data Yu and Tian ICPR 2006 Multipleclassifiers can be applied on the features generated by ADP KNN Bayesian SVM J Yu Data and Vision Seminar Interactive Boosting Step 1 Train weak classifiers on the original labeled data set and assign weights to classifiers based on their performance Step 2 Predict the labels of unlabelled data and present a subset of unlabeled data with their predicted labels to the user Step 3 User gives feedback on the retrieved data Step 4 Data obtained from user relevance feedback is added to Sonstruct a new labeled data set and removed from unlabeled ata set Step 5 The labeled data are weighted according to their predicted label correctness Step 6 Go back to Step 1 J Yu Data and Vision Seminar Interactive Boosting Pred ctlon Unlabeled Data User Feedback mter39ace Verlllcahon Data Weight Update Interactive Boosting Experiment on Benchmark Data Sets gtData set Heart and BreastCancer BC from UCI database gtPerformance of 5 iterations gtRelevance feedback of 5 data is obtained gtCompared with classic boostingADP gtClassifier KNN gtLower overall prediction error rate is obtained on both datasets 016 015 014 013 Error Rate 012 011 1 2 3 4 5 Iteration B ADP on Heart iBoost on Heart 03 028 026 Enqu 024 022 1 2 3 4 5 Iteration 3 ADP on BC iBoost on BC Yu Lu Xu and Tian ICASSP 2007 Interactive Boosting Experiment on Corel Data Sets gtData set Corel gtPerformance of 5 runs gtRelevance feedback of 5 data is obtained gtCompared with classic relevance feedback RF gtCassifier KNN gtOverall prediction precision is higher for iBoost 0 o Precision 0 00 0 1 0 ox 0 U1 Iteration I RF I iBoost Yu Lu Xu and Tian ICASSP 2007 Discussion and Conclusion Information fusion fits the nature of multimedia retrieval system Multimedia data integrates information from multimodality More robust and efficient retrieval can be achieved due to the redundant and complementary information Multiple classifier system can provide higher accuracy and handle extreme data set better Human centered computing may be a useful tool to further improve information fusion in MIR J Yu Data and Vision Seminar Extra Slides J Yu Data and Vision Seminar The Stateof theArts Single Biometric Multiple Representation Single Biometric Multiple Matcher Multiple Biometric Fusion Multimodality


Buy Material

Are you sure you want to buy this material for

25 Karma

Buy Material

BOOM! Enjoy Your Free Notes!

We've added these Notes to your profile, click here to view them now.


You're already Subscribed!

Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'

Why people love StudySoup

Jim McGreen Ohio University

"Knowing I can count on the Elite Notetaker in my class allows me to focus on what the professor is saying instead of just scribbling notes the whole time and falling behind."

Anthony Lee UC Santa Barbara

"I bought an awesome study guide, which helped me get an A in my Math 34B class this quarter!"

Steve Martinelli UC Los Angeles

"There's no way I would have passed my Organic Chemistry class this semester without the notes and study guides I got from StudySoup."

Parker Thompson 500 Startups

"It's a great way for students to improve their educational experience and it seemed like a product that everybody wants, so all the people participating are winning."

Become an Elite Notetaker and start selling your notes online!

Refund Policy


All subscriptions to StudySoup are paid in full at the time of subscribing. To change your credit card information or to cancel your subscription, go to "Edit Settings". All credit card information will be available there. If you should decide to cancel your subscription, it will continue to be valid until the next payment period, as all payments for the current period were made in advance. For special circumstances, please email


StudySoup has more than 1 million course-specific study resources to help students study smarter. If you’re having trouble finding what you’re looking for, our customer support team can help you find what you need! Feel free to contact them here:

Recurring Subscriptions: If you have canceled your recurring subscription on the day of renewal and have not downloaded any documents, you may request a refund by submitting an email to

Satisfaction Guarantee: If you’re not satisfied with your subscription, you can contact us for further help. Contact must be made within 3 business days of your subscription purchase and your refund request will be subject for review.

Please Note: Refunds can never be provided more than 30 days after the initial purchase date regardless of your activity on the site.