Text Analytics. Modeling Information Retrieval 2. Ulf Leser
|
|
- Paul Wells
- 8 years ago
- Views:
Transcription
1 Text Analytics Moeling Information etrieval 2 Ulf Leser
2 Content of this Lecture I Moels Boolean Moel Vector Sace Moel elevance Feebac in the VSM Probabilistic Moel Latent Semantic Inexing Other I Moels Ulf Leser: Text Analytics, Winter Semester 2010/2011 2
3 A Probabilistic Interretation of elevance We want to comute the robability that a oc is relevant to uery The robabilistic moel etermines this robability iteratively using user or automatic feebac Similar to VSM with relevance feebac But ifferent an more rincile way of incororating feebac Assume there is a subset D which contains all an only relevant ocuments for, an =D\ For each ocument, we want to comute the robability that belongs to for This is base on the wors in, i.e., we reresent as the set of wors containe in ={ i } Ulf Leser: Text Analytics, Winter Semester 2010/2011 3
4 Ulf Leser: Text Analytics, Winter Semester 2010/ A Probabilistic Interretation of elevance Since wors i aear both in relevant an in irrelevant ocs, we nee to loo at the influence of both We use os-scores Assuming statistical ineenence of wors, we get Clearly wrong, ~, sim rel = *...* *...*,...,,...,, ~, n n n n sim rel = =
5 Binary Ineenence Moel I Using Bayes Theorem sim, = = * * * * ~ : relative freuency of relevant irrelevant ocs in D A-Priori robability of a oc to be ir-relevant an are ineenent from an thus, both are constant for an irrelevant for raning ocuments is the robability of rawing the combination of wors forming when rawing wors at ranom from Ulf Leser: Text Analytics, Winter Semester 2010/2011 5
6 Ulf Leser: Text Analytics, Winter Semester 2010/ Binary Ineenence Moel II : Drawing the wors from means two things Drawing the wors from, an not rawing the wors not in BI consiers both lus ineenence of terms Proerties Having wors that are freuent in raises the similarity to ot having wors that are freuent in raises the similarity to Why is no in the formula? = = sim * *,
7 Ulf Leser: Text Analytics, Winter Semester 2010/ Continuation ehrasing using Focusing on the uery terms In a real setting we are not sure about an give less weight to terms not in the uery Drastically increases erformance = sim \ \ \ \ * * *, = \ \ 1 1 * *...
8 Ulf Leser: Text Analytics, Winter Semester 2010/ Last Ste Some reformulating ulicating the terms in \ 1 1 * = = 1 1 * *1 *1 1 1 * *1 *1 *1 *1 \ All uery terms All matching terms All matching terms All non-matching terms
9 Continuation 2 Obviously, the last term is ientical for all ocs. Thus sim, *1 *1 sim, = robability of a ocument comrising the terms of being relative to uery But: Comuting sim, reuires nowlege of an If an were nown, we can estimate / using maximum lielihoo This means: Comuting relative freuencies of terms in / In reality, we actually want to fin an Ulf Leser: Text Analytics, Winter Semester 2010/2011 9
10 Bac to eality Iea: Aroximation using an iterative rocess Start with some eucate guess for an set =D\ E.g. retrieve all ocs containing at least one wor from Comute robabilistic raning of all ocs wrt base on first guess Here it is imortant to focus on terms in Chose relevant ocs by user feebac or hoefully relevant ocs by selecting the to-r ocs This gives new sets an If to-r ocs are chosen, we may chose to only change robabilities of terms in an isregar the uestionable negative information Comute new term scores an new raning Iterate until satisfie [Variant of the Exectation Maximization Algorithm EM] Ulf Leser: Text Analytics, Winter Semester 2010/
11 Initialization Tyical simlifying assumtions for the start Terms in non-relevant ocs are eually istribute: ~f /D is constant, e.g., =0.5 Much less comutation, less weight to resumably unstable first values Iterations: Assume we have a new / P = {, } P = f { D, } Ulf Leser: Text Analytics, Winter Semester 2010/
12 Examle Data Text verauf haus italien gart miet blüh woll 1 Wir veraufen Häuser in Italien 2 Häuser mit Gärten zu vermieten 3 Häuser: In Italien, um Italien, um Italien herum 4 Die italienschen Gärtner sin im Garten 5 Um unser italiensches Haus blüht s Wir veraufen Blühenes 1 1 Q Wir wollen ein Haus mit Garten in Italien mieten Ulf Leser: Text Analytics, Winter Semester 2010/
13 Examle: Initialization sim, All ocs with at least on wor from ={1,2,3,4,5}, ={6} Start with initial estimations =0.5, = f /D -> verauf=blüh=2/6 Smoothing: If X=0, set X=0.01 Comute initial raning *1 *1 V H I G M B W Q sim1,= haus*1-haus*italien*1-italien / haus*1-haus*italien*1-italien =.5*1-0.01*.5* / 0.01*1-0.5*0.01*1-0.5= = 9801 sim2,= sim3,= sim4, = sim5, = 9801 sim6,= 0 Ulf Leser: Text Analytics, Winter Semester 2010/
14 Examle: Ajustment {, } P = f P = {, } D V H I G M B W Let s use the to-2 ocs as new Secon chosen arbitrarily among 1,3,4,5 ={1,2}, ={3,4,5,6} Ajust scores Q verauf=.5, verauf=2-1/6-2=1/4 haus=1~.99, haus=4-2/6-2=2/4 italien=.5, italien=4-1/6-2=3/4 gart=.5, gart=2-1/6-2=1/4 miet=.5, miet=1-1/6-2=0~0.01 Ulf Leser: Text Analytics, Winter Semester 2010/
15 Examle: e-aning sim, *1 *1 V H I G M B W Q ew raning sim1,= haus*1-haus*italien*1-italien haus*1-haus*italien*1-italien = sim2,= Ulf Leser: Text Analytics, Winter Semester 2010/
16 Pros an Cons Avantages Soun robabilistic framewor ote that VSM is strictly heuristic what is the justification for those istance measures? esults converge to most robable ocs Uner the assumtion that relevant ocs are similar by sharing term istributions that are ifferent from istributions in irrelevant ocs Disavantages First guesses often are retty ba slow convergence Terms cannot be weighte w ij {0,1} Assumes statistical ineenence of terms as most methos Has never wore convincingly better in ractice [MS07] Ulf Leser: Text Analytics, Winter Semester 2010/
17 Probabilistic Moel versus VSM with el. Feebac Publishe 1990 by Salton & Bucley Comarison base on various corora Imrovement after 1 feebac iteration Probabilistic moel BI in general worse than VSM+rel feebac IDE Probabilistic moel oes not weight terms in ocuments Probabilistic moel oes not allow to weight terms in ueries Ulf Leser: Text Analytics, Winter Semester 2010/
18 Content of this Lecture I Moels Boolean Moel Vector Sace Moel elevance Feebac in the VSM Probabilistic Moel Latent Semantic Inexing Other I Moels Ulf Leser: Text Analytics, Winter Semester 2010/
19 Latent Semantic Inexing We so-far ignore semantic relationshis between terms Homonyms: ban money, river Synonyms: House, builing, hut, villa, Hyeronyms: officer lieutenant Iea of Latent Semantic Inexing LSI Deerwester, S., Dumais, S. T., Furnas, G. W., Lanauer, T. K. an Harshman, "Inexing by latent semantic analysis." Journal of the American society for information science 416: >5000 citations! Ma many terms into fewer semantic concets Which are hien or latent in the ocs Comare ocs an uery in concet sace instea of term sace One big avantage: Can fin ocs that on t even contain the uery terms Ulf Leser: Text Analytics, Winter Semester 2010/
20 Terms an Concets Quelle: K. Aberer, I Concets are more abstract than terms Concets are more or less relate to terms an to ocs LSI fins concets automatically by matrix maniulations A concet will be a set of freuently co-occurring terms Concets from LSI cannot be selle out, but are matrix columns Ulf Leser: Text Analytics, Winter Semester 2010/
21 Term-Document Matrix Definition The term-ocument matrix M for ocs D an terms K has n=d columns an m=k rows. M[i,j]=1 iff ocument j contains term i. Wors eually well for TF or TF*IDF values Ulf Leser: Text Analytics, Winter Semester 2010/
22 Term-Document Matrix an VSM The matrix we use in VSM was a transose ocumentterm matrix =M t Having M, we can comute the vector v containing the VSM-scores of all ocs given as v=m t Ignoring score normalization Ulf Leser: Text Analytics, Winter Semester 2010/
23 What to o with a Term-Document Matrix M is not just a comfortable way of reresenting the term vectors of all ocuments M is a matrix Linear Algebra offers many ways to maniulate matrices In the following, we aroximate M by a M M shoul be smaller than M in a certain sense Less imensions; faster comutations M shoul abstract from terms to concets The less imensions cature the least freuent co-occurrences M shoul be such that M t * M t * Prouce the least error among all M of the same imension ote: We only setch LSI Ulf Leser: Text Analytics, Winter Semester 2010/
24 Term an Document Correlation M M t is calle the term correlation matrix Has K columns an K rows Similarity of terms: how often o they co-occur in a oc? M t M is calle the ocument correlation matrix Has D columns an D rows Similarity of ocs: how many terms o they share? Examle A B C 1 1 D 1 1 A B C D = A B C D A B C D M M t Term correlation matrix Ulf Leser: Text Analytics, Winter Semester 2010/
25 Some Lineare Algebra to emember Let M be a matrix The ran of M r is the maximal number of linear ineenent rows of M its imension If we have Mλ-λx=0 for x 0, then λ is calle an Eigenwert of M an x is his associate Eigenvector Eigenvectors/-werte are useful for many things In articular, one can show that a matrix M can be transforme into a iagonal matrix L with L=U -1 *M*U with U forme from the Eigenvectors of M, but only iff M has enough Eigenvectors Such L is calle similar to M; L reresents M in another vector sace, base on another basis L can be use in many cases instea of M an is easier to hanle However, our M usually will not have enough Eigenvectors Ulf Leser: Text Analytics, Winter Semester 2010/
26 Singular Value Decomosition SVD SVD is a metho to ecomose any matrix in the following way: M = X S Y t S is the iagonal matrix of the singular values of M in escening orer an has size rxr X is the matrix of Eigenvectors of M M t Y is the matrix of Eigenvectors of M t M This ecomosition is uniue an can be comute in Or 3 n=d r r n=d m=k M = X S Y t r Ulf Leser: Text Analytics, Winter Semester 2010/
27 Examle Assume for now M is uaratic an has full ran Examle for r=k=d=3 M 11 M 12 M 13 x 11 s M 21 = 0 s 22 0 M 31 M 33 x s 33 y 11 y 33 M 11 = x 11 *s 11 +x 12 *s 12 +x 13 *s 13 *y 11 + x 11 *s 21 +x 12 *s 22 +x 13 *s 23 *y 21 + x 11 *s 31 +x 12 *s 32 +x 13 *s 33 *y 31 = x 11 *s 11 *y 11 + x 12 *s 22 *y 21 + x 13 *s 33 *y 31 M 12 =... Ulf Leser: Text Analytics, Winter Semester 2010/
28 General Case M not uaratic; r < mink, All sums range from 1 to r Σ X 1i S ii Y i1 Σ X 1i S ii Y im r m=d 0 0 = n=k Σ X ni S ii Y im Σ X ni S ii Y im LSI iea: What if we sto the sums earlier, at some s<r? s ii are sorte by escening value Aggregating only over the first s s ii values catures most of M Ulf Leser: Text Analytics, Winter Semester 2010/
29 Aroximating M S can be use to aroximate M Fix some s<r; Comute M s = X s S s Y t s X s : First s columns in X S s : First s columns an first s rows in S Y s : First s rows in Y M s has the same size as M, but ifferent values For LSI, we on t nee to comute M s, but only nee X s, S s an Y s s s M s = X s S S s Y s t s Ulf Leser: Text Analytics, Winter Semester 2010/
30 s-aroximations Since the s ii are sorte in ecreasing orer The aroximation is the better, the larger s The comutation is the faster, the smaller s LSI: Only consier the to-s singular values s must be small enough to filter out noise an to rovie semantic reuction s must be large enough to reresent the iversity in the ocuments Tyical value: Otimality: After SVD, M is the matrix where M-M 2 is the smallest Ulf Leser: Text Analytics, Winter Semester 2010/
31 Geometric Interretation of SVD M is a linear transosition in a vector sace X,Y can be seen as coorination transformations, S is a linear scaling X transforms M into a vector sace where its transosition can be reresente as a linear scaling an Y transforms it bac into the vector sace of M s-aroximation M is transforme into a vector sace of lower imension such that the new imensions cature the most imortant variations in M Distances between vectors are reserve as much as ossible Universal metho: LSI has many more alications than I Ulf Leser: Text Analytics, Winter Semester 2010/
32 LI for Information etrieval We ma ocument vectors from a m-imensional sace into a s- imensional sace Aroximate ocs still are reresente by columns in Y t s Variations between ocument vectors are etermine by the number of terms they have in common The more terms in common, the smaller the istance SVD tries to reserve these istances To this en, it in a way mas freuently co-occurring terms to the same imensions Because freuently co-occurring terms have little imact on istance Freuently co-occurring terms can be interrete as concets But they cannot easily be name Also, we cannot simly etermine the terms that are mae into a new imension it is always a bit of everything a linear combination Ulf Leser: Text Analytics, Winter Semester 2010/
33 Query Evaluation After LSI, ocs are reresente by columns in Y s t How can we comute the istance between a uery an a oc in concet sace? We first nee to reresent in concet sace Assume as a new column in M Of course, we can transform M offline, but nee to transform online This woul generate a new column in Y s t To only comute this column, we aly the same transformations to as we i to all other columns of M With a little algebra, we get: = t X s S s -1 This vector is comare to the oc vectors as usual Ulf Leser: Text Analytics, Winter Semester 2010/
34 Examle: Term-Document Matrix Taen from Mi Islita: Tutorials on SVD & LSI htt:// Who too if from the Grossman an Frieer boo Query: gol silver truc M Ulf Leser: Text Analytics, Winter Semester 2010/
35 Singular Value Decomosition M = X S Y t X S Y Y t Ulf Leser: Text Analytics, Winter Semester 2010/
36 A Two-Aroximation s=2 X 2 S 2 Y 2 Y 2 t Ulf Leser: Text Analytics, Winter Semester 2010/
37 Transforming the Query = t X 2 S 2-1 Ulf Leser: Text Analytics, Winter Semester 2010/
38 Comuting the Cosine of the Angle Ulf Leser: Text Analytics, Winter Semester 2010/
39 Visualization of esults in 2D M Ulf Leser: Text Analytics, Winter Semester 2010/
40 Pros an Cons Pro Mae it into ractice, use by real search engines See-u through comutation with less imensions Increases recall an usually ecreases recision Contra Comuting SVD is exensive Fast aroximations exist, esecially for extremely sarse matrices Use stemming, sto-wor removal etc. to shrin the original matrix aning reuires less imensions than D, but more than Every uery nees to be mae first turns a few eywors into a s- imensional vector We cannot simly inex the concets of M s using inverte files etc. Thus, LSI nees other techniues than inexing rea: lots of memory Ulf Leser: Text Analytics, Winter Semester 2010/
41 Content of this Lecture I Moels Boolean Moel Vector Sace Moel elevance Feebac in the VSM Probabilistic Moel Latent Semantic Inexing Other I Moels Ulf Leser: Text Analytics, Winter Semester 2010/
42 Extene Boolean Moel One critiue to the Boolean Moel: If one term out of 10 is missing, the result is the same as if 10 were missing Iea: Measure istance for each conjunctive / isjunctive subterm of the uery exression to the ocument Examle: X-ary AD: use a rojection into x-im sace Query exression is 1,1,1,,1 Doc is a 1,a 2,,a x =0/1?,0/1?, Similarity is istance between these two oints Similar formulas for O an OT Using the aroriate efinition of istance, the extene Boolean moel may mimic both the Boolean an the VSM Ulf Leser: Text Analytics, Winter Semester 2010/
43 Generalize Vector Sace Moel One critiue to the VSM: Terms are not ineenent Thus, term vectors cannot be assume to be orthogonal Generalize Vector Sace Moel Buil a much larger vector sace with 2 K imensions Each imension minterm stans for all ocs containing a articular set of terms Minterms are not orthogonal but correlate by term co-occurrences Convert uery an ocs into minterm sace Finally, sim, is the cosine of the angel in minterm sace ice theory, inclues term co-occurrence, much more comlex than orinary VSM, no roven avantage Ulf Leser: Text Analytics, Winter Semester 2010/
An Approach to Optimizations Links Utilization in MPLS Networks
An Aroach to Otimizations Utilization in MPLS Networks M.K Huerta X. Hesselbach R.Fabregat Deartment of Telematics Engineering. Technical University of Catalonia. Jori Girona -. Camus Nor, Eif C, UPC.
More informationLatent Semantic Indexing with Selective Query Expansion Abstract Introduction
Latent Semantic Indexing with Selective Query Expansion Andy Garron April Kontostathis Department of Mathematics and Computer Science Ursinus College Collegeville PA 19426 Abstract This article describes
More informationDouble Integrals in Polar Coordinates
Double Integrals in Polar Coorinates Part : The Area Di erential in Polar Coorinates We can also aly the change of variable formula to the olar coorinate transformation x = r cos () ; y = r sin () However,
More informationMath 230.01, Fall 2012: HW 1 Solutions
Math 3., Fall : HW Solutions Problem (p.9 #). Suppose a wor is picke at ranom from this sentence. Fin: a) the chance the wor has at least letters; SOLUTION: All wors are equally likely to be chosen. The
More informationPoint Location. Preprocess a planar, polygonal subdivision for point location queries. p = (18, 11)
Point Location Prerocess a lanar, olygonal subdivision for oint location ueries. = (18, 11) Inut is a subdivision S of comlexity n, say, number of edges. uild a data structure on S so that for a uery oint
More informationThe Impact of Forecasting Methods on Bullwhip Effect in Supply Chain Management
The Imact of Forecasting Methos on Bullwhi Effect in Suly Chain Management HX Sun, YT Ren Deartment of Inustrial an Systems Engineering, National University of Singaore, Singaore Schoo of Mechanical an
More informationMeasures of distance between samples: Euclidean
4- Chapter 4 Measures of istance between samples: Eucliean We will be talking a lot about istances in this book. The concept of istance between two samples or between two variables is funamental in multivariate
More informationA MOST PROBABLE POINT-BASED METHOD FOR RELIABILITY ANALYSIS, SENSITIVITY ANALYSIS AND DESIGN OPTIMIZATION
9 th ASCE Secialty Conference on Probabilistic Mechanics and Structural Reliability PMC2004 Abstract A MOST PROBABLE POINT-BASED METHOD FOR RELIABILITY ANALYSIS, SENSITIVITY ANALYSIS AND DESIGN OPTIMIZATION
More informationSensor Network Localization from Local Connectivity : Performance Analysis for the MDS-MAP Algorithm
Sensor Network Localization from Local Connectivity : Performance Analysis for the MDS-MAP Algorithm Sewoong Oh an Anrea Montanari Electrical Engineering an Statistics Department Stanfor University, Stanfor,
More informationLecture L25-3D Rigid Body Kinematics
J. Peraire, S. Winall 16.07 Dynamics Fall 2008 Version 2.0 Lecture L25-3D Rigi Boy Kinematics In this lecture, we consier the motion of a 3D rigi boy. We shall see that in the general three-imensional
More informationPRETRIAL NEGOTIATION, LITIGATION, AND PROCEDURAL RULES
PRETRIAL NEGOTIATION, LITIGATION, AND PROCEDURAL RULES JIONG GONG AND R. PRESTON MCAFEE* We moel the ciil isute resolution rocess as a two-stage game with the arties bargaining to reach a settlement in
More informationApplication of Improved SSL in Data Security Transmission of Mobile Database System
Alication of Imrove SSL in Data Security Transmission of Mobile Database System RUIFENG WANG, XIAOHUA ZHANG, DECHAO XU College of Automation & Electrical Engineering Lanzhou Jiaotong University Lanzhou,
More information10.2 Systems of Linear Equations: Matrices
SECTION 0.2 Systems of Linear Equations: Matrices 7 0.2 Systems of Linear Equations: Matrices OBJECTIVES Write the Augmente Matrix of a System of Linear Equations 2 Write the System from the Augmente Matrix
More informationEnergy consumption in pumps friction losses
Energy consumtion in ums friction losses In this secon article in a series on energy savings in ums, Hans Vogelesang, irector of Netherlans-base esign consultancy PumSuort, eals with some ractical asects
More informationContingent Fees, Signaling and Settlement Authority
Contingent Fees, Signaling an Settlement Authority SHMUEL LESHEM USC Law School Conventional wisom suggests that uner contingent fee contracts, attorneys have an excessive incentive to settle the case;
More informationF inding the optimal, or value-maximizing, capital
Estimating Risk-Adjusted Costs of Financial Distress by Heitor Almeida, University of Illinois at Urbana-Chamaign, and Thomas Philion, New York University 1 F inding the otimal, or value-maximizing, caital
More informationComparing Dissimilarity Measures for Symbolic Data Analysis
Comaring Dissimilarity Measures for Symbolic Data Analysis Donato MALERBA, Floriana ESPOSITO, Vincenzo GIOVIALE and Valentina TAMMA Diartimento di Informatica, University of Bari Via Orabona 4 76 Bari,
More informationWavefront Sculpture Technology
Auio Engineering Society Convention Paer Presente at the th Convention 00 Setember New York, NY, USA This convention aer has been rerouce from the author's avance manuscrit, without eiting, corrections,
More informationTrap Coverage: Allowing Coverage Holes of Bounded Diameter in Wireless Sensor Networks
Tra Coverage: Allowing Coverage Holes of Boune Diameter in Wireless Sensor Networks Paul Balister Zizhan Zheng Santosh Kumar Prasun Sinha University of Memhis The Ohio State University {balistr,santosh.kumar}@memhis.eu
More informationHow To Understand The Difference Between A Bet And A Bet On A Draw Or Draw On A Market
OPTIA EXCHANGE ETTING STRATEGY FOR WIN-DRAW-OSS ARKETS Darren O Shaughnessy a,b a Ranking Software, elbourne b Corresonding author: darren@rankingsoftware.com Abstract Since the etfair betting exchange
More informationLarge-Scale IP Traceback in High-Speed Internet: Practical Techniques and Theoretical Foundation
Large-Scale IP Traceback in High-Seed Internet: Practical Techniques and Theoretical Foundation Jun Li Minho Sung Jun (Jim) Xu College of Comuting Georgia Institute of Technology {junli,mhsung,jx}@cc.gatech.edu
More informationy or f (x) to determine their nature.
Level C5 of challenge: D C5 Fining stationar points of cubic functions functions Mathematical goals Starting points Materials require Time neee To enable learners to: fin the stationar points of a cubic
More informationMetabolic control analysis in a nutshell
Metabolic control analysis in a nutshell Jan-Henrik S. Hofmeyr Det. of Biochemistry University of Stellenbosch Private Bag X, Matielan 76 Stellenbosch, South Africa jhsh@maties.sun.ac.za ABSRAC Metabolic
More informationENFORCING SAFETY PROPERTIES IN WEB APPLICATIONS USING PETRI NETS
ENFORCING SAFETY PROPERTIES IN WEB APPLICATIONS USING PETRI NETS Liviu Grigore Comuter Science Deartment University of Illinois at Chicago Chicago, IL, 60607 lgrigore@cs.uic.edu Ugo Buy Comuter Science
More informationLinear Algebra Methods for Data Mining
Linear Algebra Methods for Data Mining Saara Hyvönen, Saara.Hyvonen@cs.helsinki.fi Spring 2007 Text mining & Information Retrieval Linear Algebra Methods for Data Mining, Spring 2007, University of Helsinki
More informationDIFFRACTION AND INTERFERENCE
DIFFRACTION AND INTERFERENCE In this experiment you will emonstrate the wave nature of light by investigating how it bens aroun eges an how it interferes constructively an estructively. You will observe
More informationComplex Conjugation and Polynomial Factorization
Comlex Conjugation and Polynomial Factorization Dave L. Renfro Summer 2004 Central Michigan University I. The Remainder Theorem Let P (x) be a olynomial with comlex coe cients 1 and r be a comlex number.
More informationComputational Finance The Martingale Measure and Pricing of Derivatives
1 The Martingale Measure 1 Comutational Finance The Martingale Measure and Pricing of Derivatives 1 The Martingale Measure The Martingale measure or the Risk Neutral robabilities are a fundamental concet
More informationMemory management. Chapter 4: Memory Management. Memory hierarchy. In an ideal world. Basic memory management. Fixed partitions: multiple programs
Memory management Chater : Memory Management Part : Mechanisms for Managing Memory asic management Swaing Virtual Page relacement algorithms Modeling age relacement algorithms Design issues for aging systems
More informationWillingness to Pay for a Risk Reduction
The Economics of Climate Change C 75 Willingness to Pay for a Risk Reuction Sring 0 C Berkeley Traeger 5 Risk an ncertainty The Economics of Climate Change C 75 Back to Risk We will mostly treat the category
More informationVocabulary Problem in Internet Resource Discovery. Shih-Hao Li and Peter B. Danzig. University of Southern California. fshli, danzigg@cs.usc.
Vocabulary Problem in Internet Resource Discovery Technical Report USC-CS-94-594 Shih-Hao Li and Peter B. Danzig Computer Science Department University of Southern California Los Angeles, California 90089-0781
More informationA New Evaluation Measure for Information Retrieval Systems
A New Evaluation Measure for Information Retrieval Systems Martin Mehlitz martin.mehlitz@ai-labor.e Christian Bauckhage Deutsche Telekom Laboratories christian.bauckhage@telekom.e Jérôme Kunegis jerome.kunegis@ai-labor.e
More informationChapter 2 - Porosity PIA NMR BET
2.5 Pore tructure Measurement Alication of the Carmen-Kozeny model requires recise measurements of ore level arameters; e.g., secific surface area and tortuosity. Numerous methods have been develoed to
More informationPower analysis of static VAr compensators
Available online at www.scienceirect.com Electrical Power an Energy ystems 0 (008) 7 8 www.elsevier.com/locate/ijees Power analysis of static VAr comensators F.. Quintela *, J.M.G. Arévalo,.. eono Escuela
More informationThe Online Freeze-tag Problem
The Online Freeze-tag Problem Mikael Hammar, Bengt J. Nilsson, and Mia Persson Atus Technologies AB, IDEON, SE-3 70 Lund, Sweden mikael.hammar@atus.com School of Technology and Society, Malmö University,
More informationThe risk of using the Q heterogeneity estimator for software engineering experiments
Dieste, O., Fernández, E., García-Martínez, R., Juristo, N. 11. The risk of using the Q heterogeneity estimator for software engineering exeriments. The risk of using the Q heterogeneity estimator for
More informationImproved Algorithms for Data Visualization in Forensic DNA Analysis
Imroved Algorithms for Data Visualization in Forensic DNA Analysis Noor Maizura Mohamad Noor, Senior Member IACSIT, Mohd Iqbal akim arun, and Ahmad Faiz Ghazali Abstract DNA rofiles from forensic evidence
More informationStatic and Dynamic Properties of Small-world Connection Topologies Based on Transit-stub Networks
Static and Dynamic Proerties of Small-world Connection Toologies Based on Transit-stub Networks Carlos Aguirre Fernando Corbacho Ramón Huerta Comuter Engineering Deartment, Universidad Autónoma de Madrid,
More informationLagrangian and Hamiltonian Mechanics
Lagrangian an Hamiltonian Mechanics D.G. Simpson, Ph.D. Department of Physical Sciences an Engineering Prince George s Community College December 5, 007 Introuction In this course we have been stuying
More informationEffect Sizes Based on Means
CHAPTER 4 Effect Sizes Based on Means Introduction Raw (unstardized) mean difference D Stardized mean difference, d g Resonse ratios INTRODUCTION When the studies reort means stard deviations, the referred
More informationFailure Behavior Analysis for Reliable Distributed Embedded Systems
Failure Behavior Analysis for Reliable Distributed Embedded Systems Mario Tra, Bernd Schürmann, Torsten Tetteroo {tra schuerma tetteroo}@informatik.uni-kl.de Deartment of Comuter Science, University of
More informationOutsourcing Information Security: Contracting Issues and Security Implications
Outsourcing Information Security: Contracting Issues an Security Imlications Asunur Cezar Mile East Technical University Northern Cyress Camus Kalkanlı, Güzelyurt, KKTC, Mersin 10, Turkey asunur@metu.eu.tr
More informationAn inventory control system for spare parts at a refinery: An empirical comparison of different reorder point methods
An inventory control system for sare arts at a refinery: An emirical comarison of different reorder oint methods Eric Porras a*, Rommert Dekker b a Instituto Tecnológico y de Estudios Sueriores de Monterrey,
More informationEfficient Training of Kalman Algorithm for MIMO Channel Tracking
Efficient Training of Kalman Algorithm for MIMO Channel Tracking Emna Eitel and Joachim Seidel Institute of Telecommunications, University of Stuttgart Stuttgart, Germany Abstract In this aer, a Kalman
More informationMSc. Econ: MATHEMATICAL STATISTICS, 1995 MAXIMUM-LIKELIHOOD ESTIMATION
MAXIMUM-LIKELIHOOD ESTIMATION The General Theory of M-L Estimation In orer to erive an M-L estimator, we are boun to make an assumption about the functional form of the istribution which generates the
More informationMultiperiod Portfolio Optimization with General Transaction Costs
Multieriod Portfolio Otimization with General Transaction Costs Victor DeMiguel Deartment of Management Science and Oerations, London Business School, London NW1 4SA, UK, avmiguel@london.edu Xiaoling Mei
More informationA Multivariate Statistical Analysis of Stock Trends. Abstract
A Multivariate Statistical Analysis of Stock Trends Aril Kerby Alma College Alma, MI James Lawrence Miami University Oxford, OH Abstract Is there a method to redict the stock market? What factors determine
More informationForensic Science International
Forensic Science International 214 (2012) 33 43 Contents lists available at ScienceDirect Forensic Science International jou r nal h o me age: w ww.els evier.co m/lo c ate/fo r sc iin t A robust detection
More informationText Analytics (Text Mining)
CSE 6242 / CX 4242 Apr 3, 2014 Text Analytics (Text Mining) LSI (uses SVD), Visualization Duen Horng (Polo) Chau Georgia Tech Some lectures are partly based on materials by Professors Guy Lebanon, Jeffrey
More informationMachine Learning with Operational Costs
Journal of Machine Learning Research 14 (2013) 1989-2028 Submitted 12/11; Revised 8/12; Published 7/13 Machine Learning with Oerational Costs Theja Tulabandhula Deartment of Electrical Engineering and
More informationManaging specific risk in property portfolios
Managing secific risk in roerty ortfolios Andrew Baum, PhD University of Reading, UK Peter Struemell OPC, London, UK Contact author: Andrew Baum Deartment of Real Estate and Planning University of Reading
More informationScalar : Vector : Equal vectors : Negative vectors : Proper vector : Null Vector (Zero Vector): Parallel vectors : Antiparallel vectors :
ELEMENTS OF VECTOS 1 Scalar : physical quantity having only magnitue but not associate with any irection is calle a scalar eg: time, mass, istance, spee, work, energy, power, pressure, temperature, electric
More informationThe Fundamental Incompatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsampling
The Fundamental Incomatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsamling Michael Betancourt Deartment of Statistics, University of Warwick, Coventry, UK CV4 7A BETANAPHA@GMAI.COM Abstract
More informationHow To Find Out How To Calculate Volume Of A Sphere
Contents High-Dimensional Space. Properties of High-Dimensional Space..................... 4. The High-Dimensional Sphere......................... 5.. The Sphere an the Cube in Higher Dimensions...........
More informationCRITICAL AVIATION INFRASTRUCTURES VULNERABILITY ASSESSMENT TO TERRORIST THREATS
Review of the Air Force Academy No (23) 203 CRITICAL AVIATION INFRASTRUCTURES VULNERABILITY ASSESSMENT TO TERRORIST THREATS Cătălin CIOACĂ Henri Coandă Air Force Academy, Braşov, Romania Abstract: The
More informationThe Economics of the Cloud: Price Competition and Congestion
Submitted to Oerations Research manuscrit The Economics of the Cloud: Price Cometition and Congestion Jonatha Anselmi Basque Center for Alied Mathematics, jonatha.anselmi@gmail.com Danilo Ardagna Di. di
More informationCalculating Viscous Flow: Velocity Profiles in Rivers and Pipes
previous inex next Calculating Viscous Flow: Velocity Profiles in Rivers an Pipes Michael Fowler, UVa 9/8/1 Introuction In this lecture, we ll erive the velocity istribution for two examples of laminar
More informationThe Economics of the Cloud: Price Competition and Congestion
Submitted to Oerations Research manuscrit (Please, rovide the manuscrit number!) Authors are encouraged to submit new aers to INFORMS journals by means of a style file temlate, which includes the journal
More informationRisk in Revenue Management and Dynamic Pricing
OPERATIONS RESEARCH Vol. 56, No. 2, March Aril 2008,. 326 343 issn 0030-364X eissn 1526-5463 08 5602 0326 informs doi 10.1287/ore.1070.0438 2008 INFORMS Risk in Revenue Management and Dynamic Pricing Yuri
More informationPOISSON PROCESSES. Chapter 2. 2.1 Introduction. 2.1.1 Arrival processes
Chater 2 POISSON PROCESSES 2.1 Introduction A Poisson rocess is a simle and widely used stochastic rocess for modeling the times at which arrivals enter a system. It is in many ways the continuous-time
More informationAn important observation in supply chain management, known as the bullwhip effect,
Quantifying the Bullwhi Effect in a Simle Suly Chain: The Imact of Forecasting, Lead Times, and Information Frank Chen Zvi Drezner Jennifer K. Ryan David Simchi-Levi Decision Sciences Deartment, National
More informationCharacterizing and Modeling Network Traffic Variability
Characterizing and Modeling etwork Traffic Variability Sarat Pothuri, David W. Petr, Sohel Khan Information and Telecommunication Technology Center Electrical Engineering and Comuter Science Deartment,
More informationA Generalization of Sauer s Lemma to Classes of Large-Margin Functions
A Generalization of Sauer s Lemma to Classes of Large-Margin Functions Joel Ratsaby University College Lonon Gower Street, Lonon WC1E 6BT, Unite Kingom J.Ratsaby@cs.ucl.ac.uk, WWW home page: http://www.cs.ucl.ac.uk/staff/j.ratsaby/
More information6.042/18.062J Mathematics for Computer Science December 12, 2006 Tom Leighton and Ronitt Rubinfeld. Random Walks
6.042/8.062J Mathematics for Comuter Science December 2, 2006 Tom Leighton and Ronitt Rubinfeld Lecture Notes Random Walks Gambler s Ruin Today we re going to talk about one-dimensional random walks. In
More informationOn the (in)effectiveness of Probabilistic Marking for IP Traceback under DDoS Attacks
On the (in)effectiveness of Probabilistic Maring for IP Tracebac under DDoS Attacs Vamsi Paruchuri, Aran Durresi 2, and Ra Jain 3 University of Central Aransas, 2 Louisiana State University, 3 Washington
More informationAnswers to the Practice Problems for Test 2
Answers to the Practice Problems for Test 2 Davi Murphy. Fin f (x) if it is known that x [f(2x)] = x2. By the chain rule, x [f(2x)] = f (2x) 2, so 2f (2x) = x 2. Hence f (2x) = x 2 /2, but the lefthan
More informationTwo-resource stochastic capacity planning employing a Bayesian methodology
Journal of the Oerational Research Society (23) 54, 1198 128 r 23 Oerational Research Society Ltd. All rights reserved. 16-5682/3 $25. www.algrave-journals.com/jors Two-resource stochastic caacity lanning
More informationC-Bus Voltage Calculation
D E S I G N E R N O T E S C-Bus Voltage Calculation Designer note number: 3-12-1256 Designer: Darren Snodgrass Contact Person: Darren Snodgrass Aroved: Date: Synosis: The guidelines used by installers
More informationIntroduction to NP-Completeness Written and copyright c by Jie Wang 1
91.502 Foundations of Comuter Science 1 Introduction to Written and coyright c by Jie Wang 1 We use time-bounded (deterministic and nondeterministic) Turing machines to study comutational comlexity of
More informationCalibration of the broad band UV Radiometer
Calibration of the broa ban UV Raiometer Marian Morys an Daniel Berger Solar Light Co., Philaelphia, PA 19126 ABSTRACT Mounting concern about the ozone layer epletion an the potential ultraviolet exposure
More informationCh 10. Arithmetic Average Options and Asian Opitons
Ch 10. Arithmetic Average Options an Asian Opitons I. Asian Option an the Analytic Pricing Formula II. Binomial Tree Moel to Price Average Options III. Combination of Arithmetic Average an Reset Options
More informationTitle: Stochastic models of resource allocation for services
Title: Stochastic models of resource allocation for services Author: Ralh Badinelli,Professor, Virginia Tech, Deartment of BIT (235), Virginia Tech, Blacksburg VA 2461, USA, ralhb@vt.edu Phone : (54) 231-7688,
More informationPRIME NUMBERS AND THE RIEMANN HYPOTHESIS
PRIME NUMBERS AND THE RIEMANN HYPOTHESIS CARL ERICKSON This minicourse has two main goals. The first is to carefully define the Riemann zeta function and exlain how it is connected with the rime numbers.
More informationAchieving quality audio testing for mobile phones
Test & Measurement Achieving quality auio testing for mobile phones The auio capabilities of a cellular hanset provie the funamental interface between the user an the raio transceiver. Just as RF testing
More informationSOME PROPERTIES OF EXTENSIONS OF SMALL DEGREE OVER Q. 1. Quadratic Extensions
SOME PROPERTIES OF EXTENSIONS OF SMALL DEGREE OVER Q TREVOR ARNOLD Abstract This aer demonstrates a few characteristics of finite extensions of small degree over the rational numbers Q It comrises attemts
More informationThe fast Fourier transform method for the valuation of European style options in-the-money (ITM), at-the-money (ATM) and out-of-the-money (OTM)
Comutational and Alied Mathematics Journal 15; 1(1: 1-6 Published online January, 15 (htt://www.aascit.org/ournal/cam he fast Fourier transform method for the valuation of Euroean style otions in-the-money
More informationChapter 6. Orthogonality
6.3 Orthogonal Matrices 1 Chapter 6. Orthogonality 6.3 Orthogonal Matrices Definition 6.4. An n n matrix A is orthogonal if A T A = I. Note. We will see that the columns of an orthogonal matrix must be
More informationMODELLING OF TWO STRATEGIES IN INVENTORY CONTROL SYSTEM WITH RANDOM LEAD TIME AND DEMAND
art I. robobabilystic Moels Computer Moelling an New echnologies 27 Vol. No. 2-3 ransport an elecommunication Institute omonosova iga V-9 atvia MOEING OF WO AEGIE IN INVENOY CONO YEM WIH ANOM EA IME AN
More informationIndexing by Latent Semantic Analysis. Scott Deerwester Graduate Library School University of Chicago Chicago, IL 60637
Indexing by Latent Semantic Analysis Scott Deerwester Graduate Library School University of Chicago Chicago, IL 60637 Susan T. Dumais George W. Furnas Thomas K. Landauer Bell Communications Research 435
More informationFirewall Design: Consistency, Completeness, and Compactness
C IS COS YS TE MS Firewall Design: Consistency, Completeness, an Compactness Mohame G. Goua an Xiang-Yang Alex Liu Department of Computer Sciences The University of Texas at Austin Austin, Texas 78712-1188,
More informationSECTION 6: FIBER BUNDLES
SECTION 6: FIBER BUNDLES In this section we will introduce the interesting class o ibrations given by iber bundles. Fiber bundles lay an imortant role in many geometric contexts. For examle, the Grassmaniann
More informationNAVAL POSTGRADUATE SCHOOL THESIS
NAVAL POSTGRADUATE SCHOOL MONTEREY CALIFORNIA THESIS SYMMETRICAL RESIDUE-TO-BINARY CONVERSION ALGORITHM PIPELINED FPGA IMPLEMENTATION AND TESTING LOGIC FOR USE IN HIGH-SPEED FOLDING DIGITIZERS by Ross
More informationModelling and Resolving Software Dependencies
June 15, 2005 Abstract Many Linux istributions an other moern operating systems feature the explicit eclaration of (often complex) epenency relationships between the pieces of software
More informationALGEBRAIC SIGNATURES FOR SCALABLE WEB DATA INTEGRATION FOR ELECTRONIC COMMERCE TRANSACTIONS
ALGEBRAIC SIGNATURES FOR SCALABLE WEB DATA INTEGRATION FOR ELECTRONIC COMMERCE TRANSACTIONS Chima Adiele Deartment of Comuter Science University of Manitoba adiele@cs.umanitoba.ca Sylvanus A. Ehikioya
More informationMonitoring Frequency of Change By Li Qin
Monitoring Frequency of Change By Li Qin Abstract Control charts are widely used in rocess monitoring roblems. This aer gives a brief review of control charts for monitoring a roortion and some initial
More informationW. Heath Rushing Adsurgo LLC. Harness the Power of Text Analytics: Unstructured Data Analysis for Healthcare. Session H-1 JTCC: October 23, 2015
W. Heath Rushing Adsurgo LLC Harness the Power of Text Analytics: Unstructured Data Analysis for Healthcare Session H-1 JTCC: October 23, 2015 Outline Demonstration: Recent article on cnn.com Introduction
More informationAn Introduction to Risk Parity Hossein Kazemi
An Introduction to Risk Parity Hossein Kazemi In the aftermath of the financial crisis, investors and asset allocators have started the usual ritual of rethinking the way they aroached asset allocation
More informationtype The annotations of the 62 samples with respect to the cancer types FL, CLL, DLBCL-A, DLBCL-G.
alizadeh Samle a from a lymhoma/leukemia gene exression study Samle a for the ISIS method Format x A 2000 x 62 gene exression a matrix of log-ratio values. 2,000 genes with the highest variance across
More informationUnited Arab Emirates University College of Sciences Department of Mathematical Sciences HOMEWORK 1 SOLUTION. Section 10.1 Vectors in the Plane
United Arab Emirates University College of Sciences Deartment of Mathematical Sciences HOMEWORK 1 SOLUTION Section 10.1 Vectors in the Plane Calculus II for Engineering MATH 110 SECTION 0 CRN 510 :00 :00
More informationStock Market Value Prediction Using Neural Networks
Stock Market Value Preiction Using Neural Networks Mahi Pakaman Naeini IT & Computer Engineering Department Islamic Aza University Paran Branch e-mail: m.pakaman@ece.ut.ac.ir Hamireza Taremian Engineering
More informationAs customary, choice (a) is the correct answer in all the following problems.
PHY2049 Summer 2012 Instructor: Francisco Rojas Exam 1 As customary, choice (a) is the correct answer in all the following problems. Problem 1 A uniformly charge (thin) non-conucting ro is locate on the
More informationBranch-and-Price for Service Network Design with Asset Management Constraints
Branch-and-Price for Servicee Network Design with Asset Management Constraints Jardar Andersen Roar Grønhaug Mariellee Christiansen Teodor Gabriel Crainic December 2007 CIRRELT-2007-55 Branch-and-Price
More informationSynchronization for a DVB-T Receiver in Presence of Strong Interference
Synchronization for a DVB-T Receiver in resence of Strong Interference R. Mhiri, D. Masse, D. Schafhuber TDF-CR,, Rue Marconi, Technoôle Metz 5778 Metz Ceex 3, France Tel: +33 3 87 75 8, email: en.masse@tf.fr
More informationFactoring Variations in Natural Images with Deep Gaussian Mixture Models
Factoring Variations in Natural Images with Dee Gaussian Mixture Models Aäron van den Oord, Benjamin Schrauwen Electronics and Information Systems deartment (ELIS), Ghent University {aaron.vandenoord,
More informationRisk and Return. Sample chapter. e r t u i o p a s d f CHAPTER CONTENTS LEARNING OBJECTIVES. Chapter 7
Chater 7 Risk and Return LEARNING OBJECTIVES After studying this chater you should be able to: e r t u i o a s d f understand how return and risk are defined and measured understand the concet of risk
More informationOn the predictive content of the PPI on CPI inflation: the case of Mexico
On the redictive content of the PPI on inflation: the case of Mexico José Sidaoui, Carlos Caistrán, Daniel Chiquiar and Manuel Ramos-Francia 1 1. Introduction It would be natural to exect that shocks to
More informationBuffer Capacity Allocation: A method to QoS support on MPLS networks**
Buffer Caacity Allocation: A method to QoS suort on MPLS networks** M. K. Huerta * J. J. Padilla X. Hesselbach ϒ R. Fabregat O. Ravelo Abstract This aer describes an otimized model to suort QoS by mean
More informationOn Adaboost and Optimal Betting Strategies
On Aaboost an Optimal Betting Strategies Pasquale Malacaria 1 an Fabrizio Smerali 1 1 School of Electronic Engineering an Computer Science, Queen Mary University of Lonon, Lonon, UK Abstract We explore
More informationAN SQL EXTENSION FOR LATENT SEMANTIC ANALYSIS
Advances in Information Mining ISSN: 0975 3265 & E-ISSN: 0975 9093, Vol. 3, Issue 1, 2011, pp-19-25 Available online at http://www.bioinfo.in/contents.php?id=32 AN SQL EXTENSION FOR LATENT SEMANTIC ANALYSIS
More informationUnit 3. Elasticity Learning objectives Questions for revision: 3.1. Price elasticity of demand
Unit 3. Elasticity Learning objectives To comrehen an aly the concets of elasticity, incluing calculating: rice elasticity of eman; cross-rice elasticity of eman; income elasticity of eman; rice elasticity
More information