Com ompsci psci 201 201 Obj bjects, ects, Trade deoff offs, - PowerPoint PPT Presentation

Com ompsci psci 201 201 Obj bjects, ects, Trade deoff offs, s, NBody dy Susan Rodger January 24, 2020 1/24/2020 Compsci 201, Spring 2020 1

E is for … • Encryption • Why SSH and SSL work • Exception • A Throwable you may catch, sometimes you may rethrow 1/24/2020 Compsci 201, Spring 2020 2

Announcements • Assignment ignment P0 - grace ace perio iod to to to today 11:59p 59pm • With late penalty last change one week later • APT APT-1 1 du due e – now in grace ace perio riod d to today 11:59p 59pm • Do not accept after grace period • Discussio cussion n 3 on Januar uary 27 • Prediscussion, do before • APT APT-2 2 du due e Janua uary ry 28 • Assignmen ignment P1 du due Thursda rsday, , Jan 30 1/24/2020 Compsci 201, Spring 2020 3

When you submit an APT • Su Submit mit a REFL FLECT fo form fo for each ch APT • Submit mit REFLE LECT T fo form m fo for each ch Assignmen ignment 1/24/2020 Compsci 201, Spring 2020 4

From Last Time … Go over WOTO: Correctness Counts http://b ://bit.l it.ly/201spr y/201spring20 ing20-0122 0122-2 1/24/2020 Compsci 201, Spring 2020 5

PFTD • Objects bjects fro rom the e grou round d up • What is java.lang.Object? Its methods? • .equals(), .toString(), later .hashCode() • Conc ncep epts ts in P1: Arra rays, , Scan anners, ners, Testin ing • Completing P1 with minimal angst • Arra rayList yList fro rom high to to low level el (mos ostl tly Frid iday) • Fits into Collections hierarchy • How to build it or do it yourself: diyad 1/24/2020 Compsci 201, Spring 2020 6

Charles Isbell • Context matters: Threads • Machine learning researcher • Systems that interact intelligently with many other intelligence agents • Dean College of Computing @ gtech • Rethinking education: Online Masters in Computer Science http://www.pbs.org/newshour/bb/online-graduate-programs-offer- degrees-significant-savings/ For me, the differences are simple to state: Computationalists grok that models, languages and machines are equivalent . 1/24/2020 Compsci 201, Spring 2020 7

Algorithmic Tradeoffs • We will use e a pro roblem lem to to underst erstand and algo gorithmic rithmic trad ade-of offs and how ArrayList List works • java.util.ArrayList is "growable array", but more! • What is the class, what is the package • Package is a collection of related classes • Given en a list t of word rds, s, find the e unique ue word rds • Algorithms with ArrayLists • Alternative with Set data structure 1/24/2020 Compsci 201, Spring 2020 8

Array and ArrayList • Array can n hold primiti itive or Object ect types • int[] and String[] work • Fixed size, cannot grow • Use java. a.ut util il clas ass ArrayList fo for gro rowth, th, more https:// ps://do docs cs.o .oracle cle.co com/e m/en/ n/java/j /javase/11/do se/11/docs cs/a /api pi/j /java.ba .base/ja se/java/u /uti til/Ar l/ArrayLis yList. t.ht html ml • Contain object types, not primitives • Use .get(),.set() and not [] for indexing 1/24/2020 Compsci 201, Spring 2020 9

Look at Code: ArrayListUnique • Pro roblem lem • Read words from a file • Want the unique words in sorted order 1/24/2020 Compsci 201, Spring 2020 10

Tradeoffs: Algorithmic Approaches • https://cour urse sework. k.cs.duk uke.edu/ u/20 201sp 1spring20/cl ing20/class sscode/sr src • Read d word rds s fro rom m a file, , sto tore e in Arra rayList List, class ss is ArrayListUnique – why arra ray doesn sn't 't work? • Tradeoffs in creating sorted list of unique words • Algorithmic rithmic concep cepts ts with ArrayList List met ethod hods • Compare three different algorithmic approaches • Reasoning with and learning about Java code 1/24/2020 Compsci 201, Spring 2020 11

Method A: Add each word to a sorted list • Code e in metho ethodA: : pro roces cess s each ch word rd in list st, , add X to to list t of sorted ed, , unique que word rds • If X already in sorted-list? Nothing to do • If X greater than all words in list? Add at end • Some word greater than X? shift to make room 1/24/2020 Compsci 201, Spring 2020 12

Example: insert “egg” “egg” “egg” comes after “cat” 1/24/2020 Compsci 201, Spring 2020 13

Method A: How to shift to add "in middle" • Find first st element ement bigger ger than an String ring X at at index ex k • Shift right end to index > k, then add X there https://docs.oracle.com/en/java/javase/11/docs/api/java.base/java/util/ArrayList.html#add(int,E) 1/24/2020 Compsci 201, Spring 2020 16

Method A: Processing every string • Add unique ue element ments fro rom list t to to ret, et, keep ep sorted ed • Code reason: flag, break, best, worst cases … 1/24/2020 Compsci 201, Spring 2020 26

Method A: Details of shifting to add X • Don't 't know about ut list.add(k,X) then … • Shift from end to index > X, then add X 1/24/2020 Compsci 201, Spring 2020 27

Method B: Tradeoff: Sort first, keep unique • There e are e du duplic icates es in list, t, but it's s sorted ed • Process sorted elements, add to end if unique • Use of copy, why for-loop starts at 1 (priming) 1/24/2020 Compsci 201, Spring 2020 28

Example: Sorted with duplicates k copy 1/24/2020 Compsci 201, Spring 2020 29

Comparing Tradeoffs: Performance • Bot oth h methodA and methodB pro roces cess every ery word rd in the e list t of word rds • In loop body in methodA , shift happens • Could every element be shifted every time? • Shift 1, then 2, then 3, then … then shift N • Total work done? 1 + 2 + … + N 1/24/2020 Compsci 201, Spring 2020 38

Sort first, why is this faster? • Bot oth h metho ethodA and meth ethodB odB pro roces cess s every ery word rd in the e list t of word rds • In loop body in methodB , NO shift happens • But, all strings sorted before loop • Sorting ting takes es N x l log N fo for N strings rings • Shifti fting ng takes es 1 + 2 2 + .. + N = = N(N+1)/ )/2 • If N = = one millio lion? n? One billio ion n opera ratio tions ns/sec /second nd • Sorting is 20 million, shifting is 0.5 trillion 1/24/2020 Compsci 201, Spring 2020 39

Method C: What if we use API, other classes • A s set t contains tains no duplic licates, es, a TreeSe eeSet mainta ntains ins unique ue element ments s in sorted ed ord rder er • Create set, contains no duplicates • Create ArrayList from set • Where are the loops? 1/24/2020 Compsci 201, Spring 2020 40

What you will know … • Which ch of metho ethodA, , meth ethodB dB, , method ethodC is bet etter er? • It depends, but on what does it depend? • How does s metho ethodA scale ale as as # word rds s increa eases? es? • 1 + 2 + … + N = N(N+1)/2, just say no! • What at is log 2 (1,024)? or log 2 (1,048,576)? • Well, 2 10 = 1024 so … 1/24/2020 Compsci 201, Spring 2020 41

Why is methodA slow? • Add unique ue elements ments fro rom list t to to ret, et, keep ep sorted ed • Code reason: flag, break, best, worst cases … 1/24/2020 Compsci 201, Spring 2020 42

Java Concepts • Loops execut ecute until il loop-guar uard is false se • break exits loop early • continue re-checks guard, skipping body • Some e loops need d initia tiali lizatio tion n bef efore e loop gu guar ard • aka "priming the loop", e.g., done = false • if (!done) same e as as if (done == false) 1/24/2020 Compsci 201, Spring 2020 43

Tradeoff: Sort first, keep unique • There e are e duplic icates es in list, t, but it's s sorted ed • Process sorted elements, add to end if unique • Use of copy, why for-loop starts at 1 (priming) 1/24/2020 Compsci 201, Spring 2020 44

WOTO http://bit.l ://bit.ly/201spri y/201spring20 ng20-01 0124 24-1 1/24/2020 Compsci 201, Spring 2020 45

Measurement and Analysis • We measur sured ed runtimes ntimes empiric irically lly • Same on laptop tomorrow? Next year? • What about your computer, super computer? • Math athema ematic ical al anal alysi sis of runtim times es • Machine independent • Compare algorithms without timing them! 1/24/2020 Compsci 201, Spring 2020 46

Analysis via Pictures • Reverse erse alphab habetic tical l ord rder er, , shift ft all stri ring ngs • Shift 1, then 2, then …, finally N strings • 1+2+ … + N = N(N+1)/2 • Roughly N 2 • Square with side N? 1/24/2020 Compsci 201, Spring 2020 47

Joy Buolamwini • Found nded ed Algorithmic rithmic Justic tice e Leag ague ue • Rhodes Scholar, Anita Borg Scholar • TedX: Fighting Algorithmic Bias • Facial Recognition Bias • MIT MS with Ethan Zuckerman And so in exploring this [facial recognition], I could have viewed my face not being consistently detected as, “Oh, this is a technical challenge” — but being in the space of the Center for Civic Media definitely orients me to [say], “This is not just a technical challenge … this is as much a reflection of society as other spaces where you see inequities that need to be addressed.” https://mitadmissions.org/blogs/entry/interview-joy-buolamwini / 1/24/2020 Compsci 201, Spring 2020 48