Policy-contingent state abstraction for
|
|
- Garry Houston
- 5 years ago
- Views:
Transcription
1 Policy-contingent stte bstrction for hierrchicl MDPs oelle Pineu nd Geoffrey Gordon School of Computer Science Crnegie Mellon University Pittsburgh PA Abstrct Hierrchiclly structured plnning problems often provide gret opportunities for stte bstrction: high-level plnners cn ignore fine detils while low-level plnners cn focus only on specific tsks. Most previous hierrchicl MDP lgorithms rely on hnd-designed knowledgeintensive stte bstrctions. We propose insted n utomtic lzy lgorithm which plns from the bottom of the hierrchy up nd finds good bstrctions for high-level plnners contingent on their subtsks optiml policies. This policy-contingent pproch cn often bstrct wy more detils thn humn specificlly becuse it hs ccess to more informtion (nmely its subtsks policies) when it decides whether to cluster sttes. We demonstrte tht our lgorithm finds better bstrctions thn humn designers on well-known plnning tsks. 1 Introduction Numerous decision-mking problems hve been solved by csting them s Mrkov Decision Processes (MDPs) [12]. MDPs offer powerful frmework for discovering nytimenywhere ction selection strtegies for gents cting in stochstic domins [8]. Mny domins however require very lrge number of sttes nd/or ctions to describe the problem nd consequently re beyond the rech of trditionl MDP solution methods. In n ttempt to ddress such problems divide-nd-conquer -type pproches hve been developed tht exploit structure in the domin to fcilitte solving. Approches such s MAXQ [5] HAM [11] OPTIONS [10] nd ALisp [1] leverge structure in the control (ction & policy) spce to prtition complex tsk into mny hierrchiclly-relted subtsks. This typiclly leds to fster policy-lerning however much greter speed-up cn often be obtined by lso introducing stte bstrction s done in MAXQ nd ALisp. In this cse both the hierrchy nd stte bstrction re fixed from the beginning prior to ny problemsolving. As result the stte bstrction must be gnostic: ssuming no knowledge of locl/prtil policies nd preserving sufficient resolution to represent ny policy. This pper proposes hierrchicl MDP lgorithm which insted uses policy-contingent stte bstrction. The ide is to interleve the bstrction nd the policy determintion strting with lower-level subtsks nd moving up the hierrchy. Abstrction in higher-level subtsks tkes plce fter lower-level subtsks hve been solved thus the bstrction in
2 these high-level subtsks needs only be consistent with one policy per lower-level subtsk (s well s ll possible policies for the current subtsk). Consequently we obtin significntly more model reduction nd hence more sclble lgorithm. Other contributions of our lgorithm include the utomtiztion of the stte bstrction (in contrst with the hndcrfted clustering required in MAXQ nd ALisp) n extension to represent completion costs (s in MAXQ nd ALisp) nd finlly stright-forwrd extension to POMDPs. To see the dvntge of policy-contingent bstrction suppose tht robot wnts to nvigte round the CMU CS building strting t the first-floor entrnce nd rriving t the fourth-floor Robot Lb. Assume it hs subtsk GoToFloor(n) in which it cn use either elevtor1 (with hevy student trffic nd ending in the 4th floor lobby) or elevtor2 (with low student trffic nd ending in corridor ner the Lb). Without knowledge of GoToFloor(4) s policy the next subtsk must pln to rech the lb from two loctions. However hving first solved GoToFloor(4) which lerns to use elevtor2 then the higherlevel pln cn ignore the 4th floor lobby nd thus bstrct wy more informtion. 2 Review of MDPs A Mrkov Decision Process (MDP) is probbilistic frmework to perform optiml ction selection in stochstic domins. We ssume the stndrd formultion for MDPs [8] where n MDP is defined to be 4-tuple. A discrete set of sttes represents the domin where n gent must select between ctions in set. The gent ims to mximize its expected (discounted) sum of rewrds: "! $# &% (' *) # % where is the discount fctor nd +' )-. is the rewrd received t time / for executing in stte. Finlly the probbilistic distribution ' 10" 2 ) defines the stte-to-stte trnsition probbility conditioned on ction. The primry chllenge of MDPs is to optimize policy 3 mpping stte to ction selection such tht the expected sum of rewrds is mximized. For smll- to medium-size domins the optiml policy cn be found nlyticlly by solving the Bellmn eqution: 4 ' 5) 76 98;:< ' *)>=@? ACBEDF 3 ' ) %I 6 8 : < ' *)>=? A B DF ' 2 0 ) 4 ' 0 )HG ' 2 0 ) 4 ' 0 )HG For lrge domins however the Bellmn eqution is imprcticl nd pproximtion techniques hve been developed to reduce computtion. One importnt re of pproximtion is in using structurl ssumptions to decompose lrge problems. 2.1 Hierrchicl MDP pproches Hierrchicl MDP problem solving ddresses complex plnning problem by leverging domin knowledge to set intermedite gols. The intermedite gols define seprte subtsks nd constrin the solution serch spce thereby ccelerting solving. Existing hierrchicl MDP pproches include MAXQ [5] HAM [11] ALisp [1] nd OPTIONS [10]. Most pproches ssume tht the domin knowledge necessry to define subtsks is provided by the designer. Subtsks (we use nottion to represent subtsk) re formlly defined using combintion of elements including: llowed strt sttes expected gol sttes fixed/prtil policies reduced ction sets nd locl rewrd functions. In ddition to ccelerting problem solving dvntges of hierrchicl MDPs include the bility to: hierrchiclly relte subgols shre subtsks between multiple prents define K One exception is work on utomticlly finding intermedite gols for OPTIONS [9]. (1) (2)
3 ? subtsk-specific restricted ction spces nd stte bstrction nd finlly use either online or btch lgorithms to optimize globl policy nd lern the full vlue function. In terms of solution qulity the gurntees offered vry between hierrchicl pproches. To better discuss optimlity criterion we introduce two useful definitions (dpted from [5]): Definition 1 - Hierrchicl Optimlity: Given the clss of ll policies consistent with hierrchy then policy 3 is sid to be hierrchiclly optiml if it chieves the most rewrd mongst ll policies 3. Definition 2 - Recursive Optimlity: Given subtsk with ction set nd the clss of ll policies vilble in ssuming fixed policies for subtsks below in the hierrchy then policy 3 is sid to be recursively optiml if it chieve the most rewrd mongst ll policies 3. MAXQ chieves recursively optiml solutions wheres HAM OPTIONS nd ALisp chieve hierrchiclly optiml solutions. The min difference between the two cn be ttributed to context. A recursively optiml solution is obtined when subtsk policies re optimized without regrds to the context in which they re clled. In contrst hierrchicl optimlity is chieved by keeping trck of ll possible contexts for clling subtsks which is key when subtsks hve multiple gol sttes. There is trde-off between solution qulity nd representtion: hierrchicl optimlity is lwys t lest s good s recursive optimlity however recursively-optiml lgorithm cn llow further stte bstrction. 2.2 Automtic stte bstrction in MDPs The overll gol of stte bstrction is to infer function ' ) mpping sttes to clusters of sttes such tht we cn then perform plnning over clusters. In generl there re mny possible wys of clustering sttes. Therefore the rel chllenge lies in grouping sttes in such wy tht we cn ) pln over clusters with miniml loss of performnce compred to plnning over the entire stte spce nd b) significntly reduce plnning time. Den et l. [3] proposed simple three-step model minimiztion lgorithm for stndrd (nonhierrchicl) MDP stte bstrction. To infer ' ) function mpping sttes to the (expnding) set of clusters : Step I - Initilize stte clustering: Let ' ) ' ) if ' *) ' )( $- (3) Step II - Check stbility of ech cluster: A cluster is deemed stble iff ' ; 0 )? ' ; 0 )+ ' 5)! " (4) ACBED ACB D Step III - If cluster is unstble then split it: Let!#* $% (5) such tht Step 2 is stisfied (with corresponding re-ssignment of ' )+ "$ & ). In step I set of overly-generl clusters re proposed. Steps II nd III re then pplied itertively grdully splitting clusters ccording to slient differences in model prmeters until intr-cluster differences re sufficiently smll to ignore. This lgorithm exhibits the following desirble properties (which we stte without proof see [3 4] for detils): 1. Plnning over clusters converges to the optiml solution. 2. The lgorithm cn be relxed to llow pproximte (' -stble) stte bstrction. 3. Assuming fctorized MDP ll steps cn be implemented such tht we cn void fully enumerting the stte spce. This lst point is of prticulr interest: voiding full stte spce enumertion is prmount for lrge problems both during stte bstrction nd during plnning/solving.
4 + - Both Dietterich [5] nd Andre&Russell [1] hve shown tht there re tremendous opportunities for stte bstrction in hierrchicl MDP plnning where the prevlence of bstrction is direct result of imposing the hierrchy. In the next section we describe n lgorithm which uses utomtic stte bstrction in the context of hierrchies. 3 Hierrchicl MDPs with policy-contingent stte bstrction We propose new hierrchicl MDP pproch which interleves plnning nd stte bstrction. The motivtion behind interleved plnning nd bstrction is to dely stte bstrction for higher-level subtsks until the lower-level subtsks hve been solved. This is in contrst to the typicl pproch of first defining stte bstrction for ll subtsks nd then solving them. The min insight is tht if we dely finding the stte bstrction for subtsk until lower-level policy 3 is fixed then stte bstrction in cn be contingent only on 3 hence policy-contingent stte bstrction s opposed to llowing sufficient stte detil to ccommodte ny policy 3 (gnostic stte bstrction). Consequently we cn obtin fr more stte bstrction in higher-level subtsks. 3.1 Structurl ssumptions Our lgorithm strts with set of structurl ssumptions which re similr to erlier pproches nd more specificlly re identicl to MAXQ. Formlly we re given tsk grph (see figure 1 for n exmple) where ech lef node represents primitive ction from the originl MDP ction set. Ech internl node hs the dul role of representing both distinct subtsk (whose ction set is defined by its immedite children in the hierrchy) s well s n bstrct ction (we use br s in to denote bstrct ctions) in the context of the bove-level subtsk. A subtsk is formlly defined by: # the set of ctions which re llowed in subtsk. Bsed on the hierrchy there is one ction for ech immedite child of. the set of terminl sttes for subtsk where. Subtsk is terminted upon reching stte t which point control is returned to the higher-level subtsk which initited. ' ) the pseudo-rewrd function specifying the reltive desirbility of ech terminl stte. By definition ' 5) "$. In ddition we require prmeterized model of the MDP domin:. This is deprture from mny reinforcement lerning hierrchicl pproches however it is necessry to perform utomtic stte bstrction (see equtions 3-4.) 3.2 Algorithm Description Our lgorithm is the following: Given n MDP nd tsk hierrchy Step 1: Re-formulte structured stte spce: following bottom-up ordering: For ech subtsk Step 2: Set prmeters: ' *) nd ' *)+ "$ )+ Step 3: Minimize subtsk: Step 4: Solve subtsk: A ' 3 We use nottion for terminl sttes (both gol nd non-gol) insted of the more common in n ttempt to void confusion with the nottion for trnsition probbilities! #"$%"&('*).!.) In prctice terminl sttes. is often set to zero for gol sttes nd to lrge negtive vlue for non-gol
5 # MDP: Hierrchy: h s0 s 1 s 3 2 s h STRUCTURED STATE SPACE 3 4 h 1 s 0 h 1 s 1 h 1 s 2 h 1 s 3 h 0 s 0 h 0 s 1 h 0 s 2 h 0 s h 1 s 3 0 h 1 s 1 h 1 s 3 2 h 1 s h 1 c 3 0 h 1 c h 1 c 3 0 h 1 c 1 SET PARAMETERS for h 1 CLUSTER STATES for h 1 SOLVE POLICY for h 1 where c 0 ={s 0 s 2 c 1 ={s 1 s h 2 0 s 1 0 h 0 s 1 h 0 s 1 2 h 0 s h 2 0 c 1 0 h 0 c 1 h 0 c 2 h 0 c 1 0 h 0 c 1 SET PARAMETERS for h 0 CLUSTER STATES for h 0 SOLVE POLICY for h h 0 c 2 where c 0 ={s 0 c 1 ={s 1 c 2 ={s 2 s 3 Figure 1: Simple 4-stte problem with 2-subtsk hierrchy. Non-zero trnsition probbilities re illustrted in the MDP. The subtsk hierrchy is illustrted in the top right corner. Note tht the finl policy of K $ (bottom left) is used to model K (right column). Figure 1 illustrtes the entire process for simple 4-stte 2-subtsk problem. In step 1 Re-formulte structured stte spce we use n ide first introduced in the HAM frmework [11] tht re-formultes the MDP to reflect structurl ssumptions. The new stte spce is spnned by the cross of the originl stte spce nd hierrchy stte. Step 2 Set prmeters ppropritely trnsltes conventionl trnsition nd rewrd prmeters to the structured problem representtion. This includes copying originl trnsition nd rewrd prmeters from for ll relevnt primitive ctions s well s inferring prmeters for the5 newly-introduced bstrct ctions. Given subtsk with stte set # nd ction set where # invoke corresponding lower-level subtsks equtions 6-9 trnslte the prmeters from the originl MDP to the structured stte spce. Cse 1: Primitive ctions (see full-line trnsition rrows in figure 1) ' * ) ' * )+ * * (6) ' * ) ' )( If $ (7) ' H)+ If $ Cse 2: Abstrct ctions (see dotted-line trnsition rrows in figure 1) ' ) ' 3 ' )( )+ ' ) ' 3 ' ))( ' )( If (8) (9) One should note tht equtions 8-9 depend on 3 - the finl policy of subtsk. This enforces the policy-contingent spect of our lgorithm. Becuse prmeter setting is delyed until ll lower-level subtsks hve been solved then stte bstrction in must be sufficient to represent only 3 s opposed to ny policy 3. For the cse -! "!!)) -!!) we use true rewrd not pseudo-rewrd.
6 = Given tht ll sttes in cluster must shre the sme vlue steps 3 nd 4 cn be folded into one. Thus the model minimiztion lgorithm with vlue updtes is: Step I-III - Sme s equtions 3-5 Step IV - Updte vlue of ech cluster: 4 '!5) : < ' *)>= #? B D '!1 ; $ 0 ) 4 ' 0 )CG& "! For non-hierrchicl MDPs this is equivlent to Boutilier et l. s decision-tree representtion of MDP vlue functions [2]. ) The vlue function of the finl 4 policy solution cn be retrieved by looking t the vlue function of the top subtsk: '. Becuse it fixes low-level subtsk policies prior to solving higher-level subtsks the lgorithm is limited to recursive optimlity (rther thn hierrchicl). However for mny problems the solution qulity trde-off is smll compred to the mount of policy-contingent bstrction chieved. To execute the finl policy we dopt top-down polling pproch. At every time step we strt by querying the policy of the top subtsk nd if it returns n bstrct ction we query the policy of tht subtsk nd so on down the tree until primitive ction is returned. 3.3 Extensions to policy-contingent bstrction nd plnning A key ttribute of our lgorithm is the fct tht we model bstrct ctions ccording to their one-step effect (see equtions 8-9) s opposed to their cumultive effect: ' )? ' )+ if which most hierrchicl MDP pproches fvor (e.g. MAXQ ALisp Options). The implictions of using one-step effects re two-fold. On the positive side using one-step effects removes the need to detect subtsk terminl sttes nd therefore our lgorithm cn esily be extended to Prtilly Observble MDPs. POMDPs re generliztion of MDPs which llow for prtil stte observbility [7]. Plnning in POMDPs is typiclly exponentil in the size of the domin mking exct solutions intrctble for ll but trivil problems. To extend the policy-contingent hierrchicl lgorithm described bove to POMDPs we mke two modifictions. First we modify the model minimiztion stbility criteri to lso check for similr observtion probbilities ' 2 50E) ; nd second we set policy-contingent observtion prmeters (similrly to equtions 6&8 but for ' 5) nd ' $ 5) ). The negtive consequence of using one-step effects is tht we do not explicitly represent subtsk completion costs (s defined in MAXQ nd ALisp) which gretly reduces the opportunity for funnel bstrction. In the prgrph below we discuss n extension of our lgorithm which uses the decomposed vlue function nd chieves full funnel bstrction. *) Extension to decomposed vlue function. Given hierrchicl representtion the MDP vlue function (eqn. 1) cn be decomposed into three prts: ' ' ' )9= ' ) where is the expected rewrd of tking ction is the expected rewrd of completing subtsk fter hving tken nd is the expected rewrd of finishing the entire tsk fter hving completed. A 2-prt decomposition ws first introduced in MAXQ. The 3-prt decomposition ws lter defined in ALisp [1]. Both included hnd-crfted bstrction functions ' ) nd ' ) s well s ' ) for ALisp. (10) (11) *) = Funnel bstrction is chieved when subtsk hs few terminl sttes (for exmple moving robot to doorwy) by decoupling costs before nd fter the terminl sttes.! #"$ ) To obtin n even more concise representtion is stored only for primitive ctions nd in the cse of bstrct ctions is recursively clculted online when necessry (see [5] for detils).
7 ? ) We cn modify the policy-contingent lgorithm to utomticlly find the decomposed bstrction functions. Since policy-contingency restricts us to recursive optimlity we ignore externl costs ' ) nd compute two seprte functions: ' ) for primitive ctions (bsed strictly on immedite rewrd see eqn. 3) nd ' ) for both primitive nd bstrct ctions. In the cse of ' ) clustering requires identicl completion costs: cn be non-trivil. The min difficulty is in estimt- (see eqn. 11) nd '. 0 MAXQ discusses this in detil ) nd proposes recursive lgorithm to clculte. It is worth pointing out tht policy-contingency offers n dvntge on this issue nmely tht these must only be evluted once unlike in MAXQ which requires re-estimting for ech vlue updte. ' 0 ) ' ACB"D In prctice utomticlly finding ' ) ing ' 50) ' 0 )>= 50 ) ' ' 0 )) Extension to Q-bstrction: During model minimiztion it is lso possible to compute 4 (bstrcting over (which bstrcts over ' 5) ). This is often 4 used when hnd-crfting bstrction functions. The dvntge of bstrcting insted of is tht we cn llow different stte bstrctions under different ctions potentilly resulting in n exponentil reduction in the number of clusters. To bstrct we fix ny policy which visits every stte-ction pir nd mke new MDP whose sttes re the stte-ction pirs of nd whose trnsitions re given by our fixed policy. We then run the utomted model minimiztion lgorithm on this new MDP. n bstrction of ' ) 4 Results - Txi domin ' ) ) insted of just ' ) The txi domin is well-known hierrchicl MDP introduced by Dietterich [5]. The overll tsk (see Figure 2) is to control txi gent with the gol of picking up pssenger from n initil loction nd then dropping him/her off t desired destintion. The initil stte is selected rndomly however it is fully observble nd trnsitions re deterministic. Figure 2b represents the MAXQ control hierrchy used for this domin. The structured stte spce cn % be described by fetures: XYpssengerdestintionH where : : : :. Tble 1 compres stte clustering results for this tsk (Txi1) s well s second lrger domin (Txi2). In Txi2 the pssenger cn strt from ny loction on the grid compred to only YBRG in Txi1. This tsk is hrder for MAXQ nd ALisp becuse of the need to represent mny more completion costs in ' /. In both tsks the policy-contingent bstrction pproch (PolCA) is ble to utomticlly discover more bstrction thn either MAXQ or ALisp. #sttes #ctions QL HSMQ MAXQ ALisp PolCA-Q Txi Txi (12) Tble 1: Number of prmeters required to lern solution. QL=Q-lerning HSMQ=hierrchicl semi-mrkov Q-lerning[6] PolCA-Q=policy-contingent bstrction (clustering on Q-vlues but without the decomposed vlue extension). The bstrction results for MAXQ nd ALisp in Txi1 re published results; Txi2 results re hnd-crfted following creful reding of ech lgorithm. All lgorithms chieve optiml performnce for Txi1. PolCA-Q lerns the optiml policy for Txi2; run-time performnce of MAXQ nd ALisp is un-verified for this tsk. 5 Conclusion This pper introduces novel hierrchicl MDP lgorithm which interleves plnning nd bstrction to solve complex MDP problems. In ddition to utomting the stte bstrc-
8 Root Get Put Pickup Nv(t) Putdown North South Est West ' ) ' ) Figure 2: The txi domin is represented using four fetures: XYPssengerDestintion. The XY represent 5x5 grid world; the pssenger cn be t ny of: YBRGtxi ; the destintion is one of: YBRG. The txi gent cn select from six ctions: NSEWPickupPutdown. Actions hve uniform -1 rewrd. Rewrd for the Pickup ction is -1 when the gent is t the pssenger loction nd -10 otherwise. Rewrd for the Putdown ction is +20 when the gent is t the destintion with the pssenger in txi nd R=-10 otherwise. tion for hierrchies this lgorithm fetures previously un-exploited policy-contingent stte bstrction s well s stright-forwrd extensions to either POMDP or completion-cost representtions. We hve not yet implemented the decomposed vlue representtion for policy-contingent bstrction; this will be the subject of future work. Finlly we predict tht better understnding of the interction between ction hierrchies nd stte bstrction my led to new methods for utomticlly discovering ction hierrchies. Acknowledgements: We thnk T. Dietterich nd D. Andre for helpful exchnges concerning MAXQ nd ALisp s well s S. Thrun N.Roy nd C.Bererton for helpful comments. References [1] D. Andre nd S. Russell. Stte bstrction for progrmmble reinforcement lerning gents. In Proceedings of AAAI [2] C. Boutilier R. Derden nd M. Goldszmidt. Stochstic dynmic progrmming with fctored representtions. AI ournl [3] T. Den nd R. Givn. Model minimiztion in Mrkov decision processes. In Proceedings of the 1997 Interntionl oint Conference on Artificil Intelligence [4] T. Den R. Givn nd S. Lech. Model reduction techniques for computing pproximtely optiml solutions for Mrkov decision processes. In Proceedings of UAI [5] T. G. Dietterich. Hierrchicl reinforcement lerning with the MAXQ vlue function decomposition. ournl of Artificil Intelligence Reserch 13: [6] T. G. Dietterich. An overview of MAXQ hierrchicl reinforcement lerning. In Proceedings of the Symposium on Abstrction Reformultion nd Approximtion (SARA [7] L. P. Kelbling M. L. Littmn nd A. R. Cssndr. Plnning nd cting in prtilly observble stochstic domins. Artificil Intelligence 101: [8] L. P. Kelbling M. L. Littmn nd A. W. Moore. Reinforcement lerning: A survey. ournl of Artificil Intelligence Reserch 4: [9] A. McGovern nd A. G. Brto. Automtic discovery of subgols in reinforcement lerning using diverse density. In Proceedings of ICML [10] A. McGovern D. Precup B. Rvindrn S. Singh nd R. S. Sutton. Hierrchicl optiml control of MDPs. In Proc. of the Yle Workshop on Adptive nd Lerning Systems [11] R. Prr nd S. Russell. Reinforcement lerning with hierrchies of mchines. In Advnces in Neurl Informtion Processing Systems volume [12] R. S. Sutton nd A. G. Brto. Reinforcement Lerning: An Introduction. MIT Press 1998.
Policy-contingent state abstraction for hierarchical MDPs
Policy-contingent stte bstrction for hierrchicl MDPs Joelle Pineu nd Geoffrey Gordon School of Computer Science Crnegie Mellon University Pittsburgh, PA 15213 jpineu,ggordon@cs.cmu.edu Abstrct Hierrchiclly
More informationA New Learning Algorithm for the MAXQ Hierarchical Reinforcement Learning Method
A New Lerning Algorithm for the MAXQ Hierrchicl Reinforcement Lerning Method Frzneh Mirzzdeh 1, Bbk Behsz 2, nd Hmid Beigy 1 1 Deprtment of Computer Engineering, Shrif University of Technology, Tehrn,
More informationSolving Problems by Searching. CS 486/686: Introduction to Artificial Intelligence Winter 2016
Solving Prolems y Serching CS 486/686: Introduction to Artificil Intelligence Winter 2016 1 Introduction Serch ws one of the first topics studied in AI - Newell nd Simon (1961) Generl Prolem Solver Centrl
More informationA REINFORCEMENT LEARNING APPROACH TO SCHEDULING DUAL-ARMED CLUSTER TOOLS WITH TIME VARIATIONS
A REINFORCEMENT LEARNING APPROACH TO SCHEDULING DUAL-ARMED CLUSTER TOOLS WITH TIME VARIATIONS Ji-Eun Roh (), Te-Eog Lee (b) (),(b) Deprtment of Industril nd Systems Engineering, Kore Advnced Institute
More informationAn Efficient Divide and Conquer Algorithm for Exact Hazard Free Logic Minimization
An Efficient Divide nd Conquer Algorithm for Exct Hzrd Free Logic Minimiztion J.W.J.M. Rutten, M.R.C.M. Berkelr, C.A.J. vn Eijk, M.A.J. Kolsteren Eindhoven University of Technology Informtion nd Communiction
More informationCOMP 423 lecture 11 Jan. 28, 2008
COMP 423 lecture 11 Jn. 28, 2008 Up to now, we hve looked t how some symols in n lphet occur more frequently thn others nd how we cn sve its y using code such tht the codewords for more frequently occuring
More informationSolving Problems by Searching. CS 486/686: Introduction to Artificial Intelligence
Solving Prolems y Serching CS 486/686: Introduction to Artificil Intelligence 1 Introduction Serch ws one of the first topics studied in AI - Newell nd Simon (1961) Generl Prolem Solver Centrl component
More informationCSCI 446: Artificial Intelligence
CSCI 446: Artificil Intelligence Serch Instructor: Michele Vn Dyne [These slides were creted by Dn Klein nd Pieter Abbeel for CS188 Intro to AI t UC Berkeley. All CS188 mterils re vilble t http://i.berkeley.edu.]
More informationBall. Player X. Player O. X Goal. O Goal
Generlizing Adversril Reinforcement Lerning Willim T. B. Uther nd Mnuel M. Veloso Computer Science Deprtment Crnegie Mellon University Pittsburgh, PA 15213 futher,velosog@cs.cmu.edu Abstrct Reinforcement
More informationCSEP 573 Artificial Intelligence Winter 2016
CSEP 573 Artificil Intelligence Winter 2016 Luke Zettlemoyer Problem Spces nd Serch slides from Dn Klein, Sturt Russell, Andrew Moore, Dn Weld, Pieter Abbeel, Ali Frhdi Outline Agents tht Pln Ahed Serch
More informationDQL: A New Updating Strategy for Reinforcement Learning Based on Q-Learning
DQL: A New Updting Strtegy for Reinforcement Lerning Bsed on Q-Lerning Crlos E. Mrino 1 nd Edurdo F. Morles 2 1 Instituto Mexicno de Tecnologí del Agu, Pseo Cuhunáhuc 8532, Jiutepec, Morelos, 6255, MEXICO
More informationToday. Search Problems. Uninformed Search Methods. Depth-First Search Breadth-First Search Uniform-Cost Search
Uninformed Serch [These slides were creted by Dn Klein nd Pieter Abbeel for CS188 Intro to AI t UC Berkeley. All CS188 mterils re vilble t http://i.berkeley.edu.] Tody Serch Problems Uninformed Serch Methods
More informationAnnouncements. CS 188: Artificial Intelligence Fall Recap: Search. Today. General Tree Search. Uniform Cost. Lecture 3: A* Search 9/4/2007
CS 88: Artificil Intelligence Fll 2007 Lecture : A* Serch 9/4/2007 Dn Klein UC Berkeley Mny slides over the course dpted from either Sturt Russell or Andrew Moore Announcements Sections: New section 06:
More informationToday. CS 188: Artificial Intelligence Fall Recap: Search. Example: Pancake Problem. Example: Pancake Problem. General Tree Search.
CS 88: Artificil Intelligence Fll 00 Lecture : A* Serch 9//00 A* Serch rph Serch Tody Heuristic Design Dn Klein UC Berkeley Multiple slides from Sturt Russell or Andrew Moore Recp: Serch Exmple: Pncke
More informationCS 221: Artificial Intelligence Fall 2011
CS 221: Artificil Intelligence Fll 2011 Lecture 2: Serch (Slides from Dn Klein, with help from Sturt Russell, Andrew Moore, Teg Grenger, Peter Norvig) Problem types! Fully observble, deterministic! single-belief-stte
More informationIn the last lecture, we discussed how valid tokens may be specified by regular expressions.
LECTURE 5 Scnning SYNTAX ANALYSIS We know from our previous lectures tht the process of verifying the syntx of the progrm is performed in two stges: Scnning: Identifying nd verifying tokens in progrm.
More informationDynamic Programming. Andreas Klappenecker. [partially based on slides by Prof. Welch] Monday, September 24, 2012
Dynmic Progrmming Andres Klppenecker [prtilly bsed on slides by Prof. Welch] 1 Dynmic Progrmming Optiml substructure An optiml solution to the problem contins within it optiml solutions to subproblems.
More informationLecture 10 Evolutionary Computation: Evolution strategies and genetic programming
Lecture 10 Evolutionry Computtion: Evolution strtegies nd genetic progrmming Evolution strtegies Genetic progrmming Summry Negnevitsky, Person Eduction, 2011 1 Evolution Strtegies Another pproch to simulting
More informationMATH 25 CLASS 5 NOTES, SEP
MATH 25 CLASS 5 NOTES, SEP 30 2011 Contents 1. A brief diversion: reltively prime numbers 1 2. Lest common multiples 3 3. Finding ll solutions to x + by = c 4 Quick links to definitions/theorems Euclid
More informationIntroduction to Computer Engineering EECS 203 dickrp/eecs203/ CMOS transmission gate (TG) TG example
Introduction to Computer Engineering EECS 23 http://ziyng.eecs.northwestern.edu/ dickrp/eecs23/ CMOS trnsmission gte TG Instructor: Robert Dick Office: L477 Tech Emil: dickrp@northwestern.edu Phone: 847
More informationUnit #9 : Definite Integral Properties, Fundamental Theorem of Calculus
Unit #9 : Definite Integrl Properties, Fundmentl Theorem of Clculus Gols: Identify properties of definite integrls Define odd nd even functions, nd reltionship to integrl vlues Introduce the Fundmentl
More informationFig.25: the Role of LEX
The Lnguge for Specifying Lexicl Anlyzer We shll now study how to uild lexicl nlyzer from specifiction of tokens in the form of list of regulr expressions The discussion centers round the design of n existing
More informationComplete Coverage Path Planning of Mobile Robot Based on Dynamic Programming Algorithm Peng Zhou, Zhong-min Wang, Zhen-nan Li, Yang Li
2nd Interntionl Conference on Electronic & Mechnicl Engineering nd Informtion Technology (EMEIT-212) Complete Coverge Pth Plnning of Mobile Robot Bsed on Dynmic Progrmming Algorithm Peng Zhou, Zhong-min
More informationMidterm 2 Sample solution
Nme: Instructions Midterm 2 Smple solution CMSC 430 Introduction to Compilers Fll 2012 November 28, 2012 This exm contins 9 pges, including this one. Mke sure you hve ll the pges. Write your nme on the
More informationFile Manager Quick Reference Guide. June Prepared for the Mayo Clinic Enterprise Kahua Deployment
File Mnger Quick Reference Guide June 2018 Prepred for the Myo Clinic Enterprise Khu Deployment NVIGTION IN FILE MNGER To nvigte in File Mnger, users will mke use of the left pne to nvigte nd further pnes
More information9 Graph Cutting Procedures
9 Grph Cutting Procedures Lst clss we begn looking t how to embed rbitrry metrics into distributions of trees, nd proved the following theorem due to Brtl (1996): Theorem 9.1 (Brtl (1996)) Given metric
More informationII. THE ALGORITHM. A. Depth Map Processing
Lerning Plnr Geometric Scene Context Using Stereo Vision Pul G. Bumstrck, Bryn D. Brudevold, nd Pul D. Reynolds {pbumstrck,brynb,pulr2}@stnford.edu CS229 Finl Project Report December 15, 2006 Abstrct A
More informationIf you are at the university, either physically or via the VPN, you can download the chapters of this book as PDFs.
Lecture 5 Wlks, Trils, Pths nd Connectedness Reding: Some of the mteril in this lecture comes from Section 1.2 of Dieter Jungnickel (2008), Grphs, Networks nd Algorithms, 3rd edition, which is ville online
More informationToward Self-Referential Autonomous Learning of Object and Situation Models
Cogn Comput (2016) 8:703 719 DOI 10.1007/2559-016-9407-7 Towrd Self-eferentil Autonomous Lerning of Object nd Sitution Models Florin Dmerow 1 Andres Knobluch 2,3 Ursul Körner 3 Julin Eggert 3 Edgr Körner
More informationCS201 Discussion 10 DRAWTREE + TRIES
CS201 Discussion 10 DRAWTREE + TRIES DrwTree First instinct: recursion As very generic structure, we could tckle this problem s follows: drw(): Find the root drw(root) drw(root): Write the line for the
More informationLECT-10, S-1 FP2P08, Javed I.
A Course on Foundtions of Peer-to-Peer Systems & Applictions LECT-10, S-1 CS /799 Foundtion of Peer-to-Peer Applictions & Systems Kent Stte University Dept. of Computer Science www.cs.kent.edu/~jved/clss-p2p08
More informationHeuristic Search for Identical Payoff Bayesian Games
Heuristic Serch for Identicl Pyoff Byesin Gmes ABSTRACT Frns A. Oliehoek Informtics Institute University of Amsterdm Amsterdm The Netherlnds F.A.Oliehoek@uv.nl Jilles S. Dibngoye Lvl University Cnd University
More informationAnnouncements. CS 188: Artificial Intelligence Fall Recap: Search. Today. Example: Pancake Problem. Example: Pancake Problem
Announcements Project : erch It s live! Due 9/. trt erly nd sk questions. It s longer thn most! Need prtner? Come up fter clss or try Pizz ections: cn go to ny, ut hve priority in your own C 88: Artificil
More informationA Heuristic Approach for Discovering Reference Models by Mining Process Model Variants
A Heuristic Approch for Discovering Reference Models by Mining Process Model Vrints Chen Li 1, Mnfred Reichert 2, nd Andres Wombcher 3 1 Informtion System Group, University of Twente, The Netherlnds lic@cs.utwente.nl
More information2014 Haskell January Test Regular Expressions and Finite Automata
0 Hskell Jnury Test Regulr Expressions nd Finite Automt This test comprises four prts nd the mximum mrk is 5. Prts I, II nd III re worth 3 of the 5 mrks vilble. The 0 Hskell Progrmming Prize will be wrded
More informationOn the Detection of Step Edges in Algorithms Based on Gradient Vector Analysis
On the Detection of Step Edges in Algorithms Bsed on Grdient Vector Anlysis A. Lrr6, E. Montseny Computer Engineering Dept. Universitt Rovir i Virgili Crreter de Slou sin 43006 Trrgon, Spin Emil: lrre@etse.urv.es
More informationP(r)dr = probability of generating a random number in the interval dr near r. For this probability idea to make sense we must have
Rndom Numers nd Monte Crlo Methods Rndom Numer Methods The integrtion methods discussed so fr ll re sed upon mking polynomil pproximtions to the integrnd. Another clss of numericl methods relies upon using
More informationPPS: User Manual. Krishnendu Chatterjee, Martin Chmelik, Raghav Gupta, and Ayush Kanodia
PPS: User Mnul Krishnendu Chtterjee, Mrtin Chmelik, Rghv Gupt, nd Ayush Knodi IST Austri (Institute of Science nd Technology Austri), Klosterneuurg, Austri In this section we descrie the tool fetures,
More informationMOMDP solving algorithms comparison for safe path planning problems in urban environments
MOMDP solving lgorithms comprison for sfe pth plnning problems in urbn environments Jen-Alexis Delmer 1 nd Yoko Wtnbe 2 nd Croline P. Crvlho Chnel 3 Abstrct This pper tckles problem of UAV sfe pth plnning
More information9 4. CISC - Curriculum & Instruction Steering Committee. California County Superintendents Educational Services Association
9. CISC - Curriculum & Instruction Steering Committee The Winning EQUATION A HIGH QUALITY MATHEMATICS PROFESSIONAL DEVELOPMENT PROGRAM FOR TEACHERS IN GRADES THROUGH ALGEBRA II STRAND: NUMBER SENSE: Rtionl
More informationAI Adjacent Fields. This slide deck courtesy of Dan Klein at UC Berkeley
AI Adjcent Fields Philosophy: Logic, methods of resoning Mind s physicl system Foundtions of lerning, lnguge, rtionlity Mthemtics Forml representtion nd proof Algorithms, computtion, (un)decidility, (in)trctility
More informationECE 468/573 Midterm 1 September 28, 2012
ECE 468/573 Midterm 1 September 28, 2012 Nme:! Purdue emil:! Plese sign the following: I ffirm tht the nswers given on this test re mine nd mine lone. I did not receive help from ny person or mteril (other
More informationSOME EXAMPLES OF SUBDIVISION OF SMALL CATEGORIES
SOME EXAMPLES OF SUBDIVISION OF SMALL CATEGORIES MARCELLO DELGADO Abstrct. The purpose of this pper is to build up the bsic conceptul frmework nd underlying motivtions tht will llow us to understnd ctegoricl
More informationPointwise convergence need not behave well with respect to standard properties such as continuity.
Chpter 3 Uniform Convergence Lecture 9 Sequences of functions re of gret importnce in mny res of pure nd pplied mthemtics, nd their properties cn often be studied in the context of metric spces, s in Exmples
More informationIntroduction to Integration
Introduction to Integrtion Definite integrls of piecewise constnt functions A constnt function is function of the form Integrtion is two things t the sme time: A form of summtion. The opposite of differentition.
More informationModel-based Policy Gradient Reinforcement Learning
776 Model-bsed Policy Grdient Reinforcement Lerning Xin Wng WANGXI~CS.ORST.EDU Thoms G. Dietterlch TGD~CS.ORST.EDU Deprtment of Computer Science, Oregon Stte University, Derborn Hll 102, Corvllis, OR 97330
More informationCS321 Languages and Compiler Design I. Winter 2012 Lecture 5
CS321 Lnguges nd Compiler Design I Winter 2012 Lecture 5 1 FINITE AUTOMATA A non-deterministic finite utomton (NFA) consists of: An input lphet Σ, e.g. Σ =,. A set of sttes S, e.g. S = {1, 3, 5, 7, 11,
More information1 Introduction
Published in IET Computers & Digitl Techniques Received on 6th July 2006 Revised on 21st September 2007 ISSN 1751-8601 Hrdwre rchitecture for high-speed rel-time dynmic progrmming pplictions B. Mtthews
More information12-B FRACTIONS AND DECIMALS
-B Frctions nd Decimls. () If ll four integers were negtive, their product would be positive, nd so could not equl one of them. If ll four integers were positive, their product would be much greter thn
More informationComputing offsets of freeform curves using quadratic trigonometric splines
Computing offsets of freeform curves using qudrtic trigonometric splines JIULONG GU, JAE-DEUK YUN, YOONG-HO JUNG*, TAE-GYEONG KIM,JEONG-WOON LEE, BONG-JUN KIM School of Mechnicl Engineering Pusn Ntionl
More informationMA1008. Calculus and Linear Algebra for Engineers. Course Notes for Section B. Stephen Wills. Department of Mathematics. University College Cork
MA1008 Clculus nd Liner Algebr for Engineers Course Notes for Section B Stephen Wills Deprtment of Mthemtics University College Cork s.wills@ucc.ie http://euclid.ucc.ie/pges/stff/wills/teching/m1008/ma1008.html
More informationA Scalable and Reliable Mobile Agent Computation Model
A Sclble nd Relible Mobile Agent Computtion Model Yong Liu, Congfu Xu, Zhohui Wu, nd Yunhe Pn College of Computer Science, Zhejing University Hngzhou 310027, Chin cckffe@yhoo.com.cn Abstrct. This pper
More informationAn introduction to model checking
An introduction to model checking Slide 1 University of Albert Edmonton July 3rd, 2002 Guy Trembly Dépt d informtique UQAM Outline Wht re forml specifiction nd verifiction methods? Slide 2 Wht is model
More information4452 Mathematical Modeling Lecture 4: Lagrange Multipliers
Mth Modeling Lecture 4: Lgrnge Multipliers Pge 4452 Mthemticl Modeling Lecture 4: Lgrnge Multipliers Lgrnge multipliers re high powered mthemticl technique to find the mximum nd minimum of multidimensionl
More informationTopic: Software Model Checking via Counter-Example Guided Abstraction Refinement. Having a BLAST with SLAM. Combining Strengths. SLAM Overview SLAM
Hving BLAST with SLAM Topic: Softwre Model Checking vi Counter-Exmple Guided Abstrction Refinement There re esily two dozen SLAM/BLAST/MAGIC ppers; I will skim. # # Theorem Proving Combining Strengths
More informationSlides for Data Mining by I. H. Witten and E. Frank
Slides for Dt Mining y I. H. Witten nd E. Frnk Simplicity first Simple lgorithms often work very well! There re mny kinds of simple structure, eg: One ttriute does ll the work All ttriutes contriute eqully
More informationEnginner To Engineer Note
Technicl Notes on using Anlog Devices DSP components nd development tools from the DSP Division Phone: (800) ANALOG-D, FAX: (781) 461-3010, EMAIL: dsp_pplictions@nlog.com, FTP: ftp.nlog.com Using n ADSP-2181
More informationSemistructured Data Management Part 2 - Graph Databases
Semistructured Dt Mngement Prt 2 - Grph Dtbses 2003/4, Krl Aberer, EPFL-SSC, Lbortoire de systèmes d'informtions réprtis Semi-structured Dt - 1 1 Tody's Questions 1. Schems for Semi-structured Dt 2. Grph
More informationChapter 2 Sensitivity Analysis: Differential Calculus of Models
Chpter 2 Sensitivity Anlysis: Differentil Clculus of Models Abstrct Models in remote sensing nd in science nd engineering, in generl re, essentilly, functions of discrete model input prmeters, nd/or functionls
More informationThe Search for Optimality in Automated Intrusion Response
The Serch for Optimlity in Automted Intrusion Response Yu-Sung Wu nd Surbh Bgchi (DCSL) & The Center for Eduction nd Reserch in Informtion Assurnce nd Security (CERIAS) School of Electricl nd Computer
More informationAddress Register Assignment for Reducing Code Size
Address Register Assignment for Reducing Code Size M. Kndemir 1, M.J. Irwin 1, G. Chen 1, nd J. Rmnujm 2 1 CSE Deprtment Pennsylvni Stte University University Prk, PA 16802 {kndemir,mji,guilchen}@cse.psu.edu
More informationINTRODUCTION TO SIMPLICIAL COMPLEXES
INTRODUCTION TO SIMPLICIAL COMPLEXES CASEY KELLEHER AND ALESSANDRA PANTANO 0.1. Introduction. In this ctivity set we re going to introduce notion from Algebric Topology clled simplicil homology. The min
More informationDigital Design. Chapter 6: Optimizations and Tradeoffs
Digitl Design Chpter 6: Optimiztions nd Trdeoffs Slides to ccompny the tetbook Digitl Design, with RTL Design, VHDL, nd Verilog, 2nd Edition, by Frnk Vhid, John Wiley nd Sons Publishers, 2. http://www.ddvhid.com
More informationPresentation Martin Randers
Presenttion Mrtin Rnders Outline Introduction Algorithms Implementtion nd experiments Memory consumption Summry Introduction Introduction Evolution of species cn e modelled in trees Trees consist of nodes
More informationAlignment of Long Sequences. BMI/CS Spring 2012 Colin Dewey
Alignment of Long Sequences BMI/CS 776 www.biostt.wisc.edu/bmi776/ Spring 2012 Colin Dewey cdewey@biostt.wisc.edu Gols for Lecture the key concepts to understnd re the following how lrge-scle lignment
More informationSmall Business Networking
Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd processes. Introducing technology
More informationCS143 Handout 07 Summer 2011 June 24 th, 2011 Written Set 1: Lexical Analysis
CS143 Hndout 07 Summer 2011 June 24 th, 2011 Written Set 1: Lexicl Anlysis In this first written ssignment, you'll get the chnce to ply round with the vrious constructions tht come up when doing lexicl
More informationStained Glass Design. Teaching Goals:
Stined Glss Design Time required 45-90 minutes Teching Gols: 1. Students pply grphic methods to design vrious shpes on the plne.. Students pply geometric trnsformtions of grphs of functions in order to
More informationA Tautology Checker loosely related to Stålmarck s Algorithm by Martin Richards
A Tutology Checker loosely relted to Stålmrck s Algorithm y Mrtin Richrds mr@cl.cm.c.uk http://www.cl.cm.c.uk/users/mr/ University Computer Lortory New Museum Site Pemroke Street Cmridge, CB2 3QG Mrtin
More informationEngineer To Engineer Note
Engineer To Engineer Note EE-186 Technicl Notes on using Anlog Devices' DSP components nd development tools Contct our technicl support by phone: (800) ANALOG-D or e-mil: dsp.support@nlog.com Or visit
More informationUSING HOUGH TRANSFORM IN LINE EXTRACTION
Stylinidis, Efstrtios USING HOUGH TRANSFORM IN LINE EXTRACTION Efstrtios STYLIANIDIS, Petros PATIAS The Aristotle University of Thessloniki, Deprtment of Cdstre Photogrmmetry nd Crtogrphy Univ. Box 473,
More informationSection 3.1: Sequences and Series
Section.: Sequences d Series Sequences Let s strt out with the definition of sequence: sequence: ordered list of numbers, often with definite pttern Recll tht in set, order doesn t mtter so this is one
More informationCMSC 331 First Midterm Exam
0 00/ 1 20/ 2 05/ 3 15/ 4 15/ 5 15/ 6 20/ 7 30/ 8 30/ 150/ 331 First Midterm Exm 7 October 2003 CMC 331 First Midterm Exm Nme: mple Answers tudent ID#: You will hve seventy-five (75) minutes to complete
More information2 Computing all Intersections of a Set of Segments Line Segment Intersection
15-451/651: Design & Anlysis of Algorithms Novemer 14, 2016 Lecture #21 Sweep-Line nd Segment Intersection lst chnged: Novemer 8, 2017 1 Preliminries The sweep-line prdigm is very powerful lgorithmic design
More informationScanner Termination. Multi Character Lookahead. to its physical end. Most parsers require an end of file token. Lex and Jlex automatically create an
Scnner Termintion A scnner reds input chrcters nd prtitions them into tokens. Wht hppens when the end of the input file is reched? It my be useful to crete n Eof pseudo-chrcter when this occurs. In Jv,
More informationUNIT 11. Query Optimization
UNIT Query Optimiztion Contents Introduction to Query Optimiztion 2 The Optimiztion Process: An Overview 3 Optimiztion in System R 4 Optimiztion in INGRES 5 Implementing the Join Opertors Wei-Png Yng,
More informationSmall Business Networking
Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd business. Introducing technology
More informationA Formalism for Functionality Preserving System Level Transformations
A Formlism for Functionlity Preserving System Level Trnsformtions Smr Abdi Dniel Gjski Center for Embedded Computer Systems UC Irvine Center for Embedded Computer Systems UC Irvine Irvine, CA 92697 Irvine,
More informationCHAPTER III IMAGE DEWARPING (CALIBRATION) PROCEDURE
CHAPTER III IMAGE DEWARPING (CALIBRATION) PROCEDURE 3.1 Scheimpflug Configurtion nd Perspective Distortion Scheimpflug criterion were found out to be the best lyout configurtion for Stereoscopic PIV, becuse
More informationDetermining Single Connectivity in Directed Graphs
Determining Single Connectivity in Directed Grphs Adm L. Buchsbum 1 Mrtin C. Crlisle 2 Reserch Report CS-TR-390-92 September 1992 Abstrct In this pper, we consider the problem of determining whether or
More informationComputer-Aided Multiscale Modelling for Chemical Process Engineering
17 th Europen Symposium on Computer Aided Process Engineesing ESCAPE17 V. Plesu nd P.S. Agchi (Editors) 2007 Elsevier B.V. All rights reserved. 1 Computer-Aided Multiscle Modelling for Chemicl Process
More informationIf f(x, y) is a surface that lies above r(t), we can think about the area between the surface and the curve.
Line Integrls The ide of line integrl is very similr to tht of single integrls. If the function f(x) is bove the x-xis on the intervl [, b], then the integrl of f(x) over [, b] is the re under f over the
More informationCS 268: IP Multicast Routing
Motivtion CS 268: IP Multicst Routing Ion Stoic April 5, 2004 Mny pplictions requires one-to-mny communiction - E.g., video/udio conferencing, news dissemintion, file updtes, etc. Using unicst to replicte
More informationWhat are suffix trees?
Suffix Trees 1 Wht re suffix trees? Allow lgorithm designers to store very lrge mount of informtion out strings while still keeping within liner spce Allow users to serch for new strings in the originl
More informationDIFFERENT POLICY OBJECTIVES OF THE ROAD AUTHORITY IN THE OPTIMAL TOLL DESIGN PROBLEM
DIFFERENT POLICY OBJECTIVES OF THE ROAD AUTHORITY IN THE OPTIMAL TOLL DESIGN PROBLEM Dusic Joksimovic*, Michiel Bliemer Delft University of Technology, The Netherlnds ARS Trnsport &Technology, The Netherlnds
More informationStatistical classification of spatial relationships among mathematical symbols
2009 10th Interntionl Conference on Document Anlysis nd Recognition Sttisticl clssifiction of sptil reltionships mong mthemticl symbols Wl Aly, Seiichi Uchid Deprtment of Intelligent Systems, Kyushu University
More informationSmall Business Networking
Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd processes. Introducing technology
More informationLooking up objects in Pastry
Review: Pstry routing tbles 0 1 2 3 4 7 8 9 b c d e f 0 1 2 3 4 7 8 9 b c d e f 0 1 2 3 4 7 8 9 b c d e f 0 2 3 4 7 8 9 b c d e f Row0 Row 1 Row 2 Row 3 Routing tble of node with ID i =1fc s - For ech
More informationParallel Square and Cube Computations
Prllel Squre nd Cube Computtions Albert A. Liddicot nd Michel J. Flynn Computer Systems Lbortory, Deprtment of Electricl Engineering Stnford University Gtes Building 5 Serr Mll, Stnford, CA 945, USA liddicot@stnford.edu
More informationMisrepresentation of Preferences
Misrepresenttion of Preferences Gicomo Bonnno Deprtment of Economics, University of Cliforni, Dvis, USA gfbonnno@ucdvis.edu Socil choice functions Arrow s theorem sys tht it is not possible to extrct from
More informationApproximation by NURBS with free knots
pproximtion by NURBS with free knots M Rndrinrivony G Brunnett echnicl University of Chemnitz Fculty of Computer Science Computer Grphics nd Visuliztion Strße der Ntionen 6 97 Chemnitz Germny Emil: mhrvo@informtiktu-chemnitzde
More informationIntegration. October 25, 2016
Integrtion October 5, 6 Introduction We hve lerned in previous chpter on how to do the differentition. It is conventionl in mthemtics tht we re supposed to lern bout the integrtion s well. As you my hve
More informationsuch that the S i cover S, or equivalently S
MATH 55 Triple Integrls Fll 16 1. Definition Given solid in spce, prtition of consists of finite set of solis = { 1,, n } such tht the i cover, or equivlently n i. Furthermore, for ech i, intersects i
More informationAlgorithm Design (5) Text Search
Algorithm Design (5) Text Serch Tkshi Chikym School of Engineering The University of Tokyo Text Serch Find sustring tht mtches the given key string in text dt of lrge mount Key string: chr x[m] Text Dt:
More information6.2 Volumes of Revolution: The Disk Method
mth ppliction: volumes by disks: volume prt ii 6 6 Volumes of Revolution: The Disk Method One of the simplest pplictions of integrtion (Theorem 6) nd the ccumultion process is to determine so-clled volumes
More informationCone Cluster Labeling for Support Vector Clustering
Cone Cluster Lbeling for Support Vector Clustering Sei-Hyung Lee Deprtment of Computer Science University of Msschusetts Lowell MA 1854, U.S.A. slee@cs.uml.edu Kren M. Dniels Deprtment of Computer Science
More informationFig.1. Let a source of monochromatic light be incident on a slit of finite width a, as shown in Fig. 1.
Answer on Question #5692, Physics, Optics Stte slient fetures of single slit Frunhofer diffrction pttern. The slit is verticl nd illuminted by point source. Also, obtin n expression for intensity distribution
More informationCS481: Bioinformatics Algorithms
CS481: Bioinformtics Algorithms Cn Alkn EA509 clkn@cs.ilkent.edu.tr http://www.cs.ilkent.edu.tr/~clkn/teching/cs481/ EXACT STRING MATCHING Fingerprint ide Assume: We cn compute fingerprint f(p) of P in
More informationx )Scales are the reciprocal of each other. e
9. Reciprocls A Complete Slide Rule Mnul - eville W Young Chpter 9 Further Applictions of the LL scles The LL (e x ) scles nd the corresponding LL 0 (e -x or Exmple : 0.244 4.. Set the hir line over 4.
More informationNetwork Interconnection: Bridging CS 571 Fall Kenneth L. Calvert All rights reserved
Network Interconnection: Bridging CS 57 Fll 6 6 Kenneth L. Clvert All rights reserved The Prolem We know how to uild (rodcst) LANs Wnt to connect severl LANs together to overcome scling limits Recll: speed
More information