MANN ECM5901 Optimization Theory and Application Spring 2022 Institute of Communications Engineering National Yang Ming Chiao Tung University Instructor: Office: Office hours: email: Prof. Ching-Yi Lai ED926 Wednesday 9:50-11:00am (or by appointment) cylai@nycu.edu.tw Lectures: 2BCD-EC015 TA: √Z office:ED717 office hour:Thursday 13:30-15:00 email:xdlaughhaha.eed06@nctu.edu.tw Text: Convex Optimization (Cambridge, 2004) Stephen Boyd and Lieven Vandenberghe Reference: 1. CVX: Matlab software http://cvxr.com/cvx/ 2. Convex Optimization for Signal Processing and Communications: From Fundamentals to Applications (CRC Press, 2017) Chong-yung Chi, Wei-chiang Li, Chia-hsiang Lin Course grade: Homework (25%): Midterm Exam (35%): Final Exam (40%): Required preparation: seven to eight problem sets will be assigned at 1-2 week intervals April 12, Tuesday 9:00-12:00 May 31, Tuesday 9:00-12:00 允 % a basic understanding of probability theory and advanced analysis and a good knowledge of linear algebra 1 鹽 Learning Objectives _ nu smpkxdgn onvexfyconcae Convex optimization is a class of mathematical optimization problems, which includes least-square and linear programming problems that can be efficiently solved. The idea is to extend known tools for these two problems to a larger class of convex optimization problems, such as semidefinite programming. SDP Convex optimization has found applications in many areas, such as estimation and signal processing, communications and networks, machine learning, circuit design, analysis, statistics, and finance, where it is used to find optimal or approximate solutions. The goal of this course is to develop a working knowledge for recognizing or formulating a problem as a convex optimization problem, which can then be reliably and efficiently solved by well developed numerical analysis tools. " 火㼉 " 焮 * maxfnmatnxst.constántsc , a The following materials in [HV03] will be covered in this semester: Chapter 1: Introduction Chapter 2: convex sets Appendx A. Chapter 3: convex functions Chapter 4: convex optimization problems - - Midtrm Chapter 5: duality theorem Chapters 9, 10, 11: Algorithms Applications of convex optimization could be found in Chapters 6, 7, and 8, which will be skipped due to the time-limit. Using CVX (Matlab software) http://cvxr.com/cvx/ 1. Download. Example: download cvx-w64.zip if you are using Windows. Then unzip the file. 2. Execute MATLAB. Run cvx setup.m and cvx startup.m in MATLAB. 3. Start to program. 2 M-PT.EZ.ae/U,fndPorq.? ※恂 Assume Faang 3.* ⼆ 百 , " 2 < M- 2 → " - nodassīalalyithm andotomgmpolgnomidtne.io mndtmgantmaynunn亝 - 、 Chapl ⼈ 正在和在 Amhhdopàmoaānpnobkn frm.variabkXDXERnrealnumbe.rs " 似 在 n-dnsmdulhs@xEldicompanumbers3xERM.mxn xīn R 成 : realmn.es XECT.icompkx.opnvefuntmfiEIR.constdfm.tn ④ f ☒) : i-y.c.im minimize.fi/sguTtofM1-EEE.IE EbisEl.2. m-E E HExfx):lR-snTfixi=xi$ ( ✗ Onxn cnnynumt) Igatgi , Da.be Riazbsa-b.a.be 成 感 " 比 an Qxb 關 QDDi.lbz.cn/an3bn.taa-NiB?wndfned. 件 3 " ( bij ) " Afijdni.c.am 川 "" " A.BE R " B A3B-AB30.ie A , - - = Bislposhesemig Teeigenváuesof (A) arenonnghe Mh MMT 店 了店 可 Mìsnormalrf brn - 2 wneretdaggerjmeanstnitiaviugāletranspose nconjugat.co Mignǖ Mhasaspedrumidwmpoat.me Cspdd Theaan) A matùx Tfandonyif M ⼆点 Zìviǘ , wnerexi.viareeigen-pa.rs Viìsaneigenvectr Mviexivi 啊 , efn Amatrìx Mispone-semidefirite. TN 䫾) E 㖄ㄨ 3 , 0 TMX Anfnǎhix 30 txtli Uxecn Mispne-semidefiteltandong.f-thsg. g e eigenuluesarenon Chomew啊 , Nesay 我 :D → Rìsconuìf Cwncàve) faxtpg > ← (3) xfxtpfy D.x.pt/RDEX.p,E1cxtp-l wnerex.ge ※ f. 八 " 的 機邁 秘 , 嫌 找 tpgtgxtgy ( convex) Nltpfy 八 xxtpy , f lminināhi ⼀ 年 : wncave , , isoptmdtfgaf.cz ) ftabi.F-c.im/Ex:XElRfx)-ax4tnPtcxEd兩 xte川fndmaxf1 nstorderant. UZ sāhsfyīng Lsaa ⼀比 7 ✗ best-sgarepnifn nsnn.ie 哈 㶯 wnareflf.be Pi (我成 ( Axb) 和 = - , ☒ 肛 啊 ( Axb ) Axttib) 煳「 datattnganencx.fi Eample = " + 凸 : ) Gf 灲 。 , , … Gkjfn) R-PifndcxsuchtmtIH-fipsmimized.0T-lbnt r.fge xiEIR.fi wrìabks a. b. , anear 4) Prgram g Ex mnimīze subjutto ← Axtb mxn mxl " wwe sohu.n(ii) o XEIR aamie.in -2 " 㓹 ⼈ 之 紅 lnearprogram max xtxz subjǜto " + ft x 30 炒0 5 Xzt 5 Zxtxz 不 ol ☒我 2 2 灶如巧 E 2 harpngm dfweanformalatapmtdpoblanasaconrexprobkm.tn sogf ph omomph resourceispolgnomialmzlchanlengei.T itcan 名 glig & 臥 refrmhgthepwblan chysherapproihmprhn ( / aix-biDCnk-no.ve ⼀班 rx 靠 GER ! bi ER hoducevaùabk TER ts.t.sgt.to/aFx-bikt Fl,. .,k.ysuunthM wūnimne ⼀⼀ ⽣ lalEbaiTx-biEtaEb-@x-b.taebYEt Tisisalnarprogram # 。