Algorithms, programming * Algorithm: how to systematically perform a task * Write down as a sequence of steps * “Recipe”, or program * Programming language describes the steps * What is a step? Degrees of detail * “Arrange the chairs” vs “Make 8 rows with 10 chairs in each row"Our focus + Algorithms that manipulate information * Compute numencal functions — f(x,y) = x’ * Reorganize data — arrange in ascending order * Optimization — find the shortest route * And more ... * Solve Sudoku, play chess, correct spelling ...Greatest common divisor * gcd(m,n) * Largest k such that k divides m and k divides n * gcd(8,12) = 4 * gcd(18,25) = 1 * 1 divides every number * At least one common divisor for every m,nComputing gcd(m, n) * List out factors of m * List out factors of n * Report the largest number that appears on both lists * Is this a valid algorithm? * Finite presentation of the “recipe” * Terminates after a finite number of stepsComputing gcd(m, n) * Factors of m must be between 1 and m * Test each number in this range * If it divides m without a remainder, add it to list of factors * Example: gcd( 14,63) * Factors of 14 12 3 4 5 6 7 8 9 1011 12 13 14Computing gcd(14, 63) * Factors of 14 12 7 44 * Factors of 63 13 7 9 2163 * Construct list of common factors * For each factor Co if it is a factor of 63 * Return largest factor in this list: 7 aeComputing gcd(14, 63) * Factors of 14 12 7 14 * Factors of 63 1 3 7 9 21 6 * Construct list of common factors * For each factor epi ff it is a factor of 63 * Return largest factor in this list! 7 aeAn algorithm for gcd(m,n_ 8 « Use fm, fn for list of factors of m, n, respectively * For each i from 1 to m, add 1 to fm if i divides m * For each j from 1 ton, add j to fn if ) divides n * Use cf for list of common factors [12,44] [13,344.63] * For each f in fim, add f to cf if f also appeafs in fnAn algorithm for gcd(m,n) * Use fm, fr for list of factors of m, n, respectively * For each i from 1 to m, add i to fm if i divides m * For each j from 1 ton, add 3 to fn if j divides n * Use cf for list of common factors * For each f in fm, add f to cf if f also appears in fn * Return largest (rightmost) value in cfOur first Python program def gcd(m,n): fa - 0 for i in range(1 ool): if (Ki) == 0: fm. append( i) fn- (] for j tn range(1,n+1) if (rij) = @: fn. append(}) ef - (1) for f in fm: if f in fn: cf .append(f) return(cf[-1])Some points to note « Use names to remember intermediate values em, n, fm, fn, cf, i, j, f * Values can be single items or collections “m,n, i, j, f are single numbers * fm, fn, cf are lists of numbers * Assign values to names * Explicitly, fn = (J, and implicitly, for f in cf: * Update them, fn.append(i)Some points to note ... * Program is a sequence of steps * Some steps are repeated * Do the same thing for each item in a list + Some steps are executed conditionally * Do something if a value meets some requirement if (mu) --0An algorithm for.gcd(m, n) + Use fm, fn for list of factors of m, n, respectively * For each i from 1 tom, add 1 to fm if 1 divides m * For each ) from 1 ton. add j to fn if j divides n * Use cf for list of common factors * For each f in fm, add f to cf if f also appears in fn * Return largest (rightmost) value in cfCan we do better? * We scan from 1 to m to compute fm and again from 1 to n to compute fn * Why not a single scan from 1 to max(m,n)? * For each i in 1 to max(m,n), add i to fmifi divides m and add i to fn if 1 divides nEven better? * Why compute two lists and then compare them to compute common factors cf? Do it in one shot. * For each i in 1 to max(m,n), if t divides m and i also divides n, then add 1 to cf * Actually, any common factor must be less than min(m,n) * For each 1 in 1 to min(m,n), if . divides m and i also divides n, then add 1 to cfA shorter Python program e def gcd(m,n): f= for i in range(1,min(m,n)+1): if (mi) == @ and (n%i) == 0: cf .append(i) return(cf[-1])A shorter Python program def gcd(m,n): ~@f=D for i in range(1,min(m,n}+1): if (mi) == @ and (n%i) == @: cf -append(i) return(cf[-1))Do we need lists at all? * We only need the largest common factor * 1 will always be a common factor * Each time we find a larger common factor, discard the previous one * Remember the largest common factor seen so far and return it * mrcf — most recent common factorNo lists! def gcd(m,n): for i in rengeCi,mintn n+): if (Wi) == @ and (n#i) <= @: mrcf = i return(mrcf)Scan backwards? * To find the largest common factor, start at the end and work backwards * Let i run from min(m,n) to 1 * First common factor that we find will be gcd!No lists! - oo trv im in cage (| me (madd) fy def gcd(m,n): i = min(m,n) Cwhile>j> @: LF (néi) == @ and (n%i) == 0: | return(i)h— bx else: lei-lA new kind of repetition while condition: ~*\step 1 “istep 2 j.- - step k * Don't know in advance how many times we will repeat the steps * Should be careful to ensure the loop terminates — eventually the condition should become faise!Summary * With a little thought, we have dramatically simplified our naive algonthm * Though the newer versions are simpler, they still take time proportional to the values m and n * Amuch more efficient approach is possibleAlgorithm for gcdCm,n) * To find the largest common factor, start at the end and work backwards * Let i run from min(m,n) to 1 * First common factor that we find will be gcd!Euclid’s algorithm * Consider gcd(m,n) with m>n * If n divides m, return n * Otherwise, compute gcd(n,m-n) and return that valueEuclid’s algorithm def gcd(m,n): # Assume m >= n ifmen: (m,n) = (n,m) if (mtn) == @: return(n) else: diff = m-n # diff > n? Possible! return(gcd(max(n, diff) ,min(n,diff))Euclid’s algorithm =! °} > wY J def gcd(m,n): \ Lu (#) Assume m >= n. Con ne’ rm t ifmen: ¢) (n,n) = (nym Miym v * R= Sy Yr sn if (mn) == 0: m2 92 return(n) A / y ne? else: v Gs C1 ee 4 de diff = m-n : ALh _ as # diff > n? Possible! Leiria return(gcd(max(n, diff) ,min(n, diff) —— esEuclid’s algorithm, again def gcd(m,n): if m= n (m,n) = (n,m) while (mn) != @: diff = m-n # diff > n? Possible! (m,n) = Cmax(n, diff) ,min(n, dif f)) return(n)Euclid’s algorithm, again def gcd(m,n): - — > ifm=n (m,n) = (nym) while (tin) (1=)0: # “= . diff = m-n # diff > n? Possible! Caen - (max(n, diff) ,min(n, diff) return(n)Even better * Suppose n does not divide = * Thenm « qn + r, where q is the quotient, r is the remainder when we divide m by n * Assume d divides both m and n * Thenm = ad,n = bdEven better + Suppose n does not divide = + fr, where q is the quotient, r is the iad ler when we divide m by n * Assume fi divities both m and n " mitts n alm * Soad = Ye oatzolanat,Euclid’s algorithm, revisited def gcd(m,n): if m= n (m,n) = (n,m) while (nn) t= 0: (m,n) = (nmin) # mn < n, always! return(n)Efficiency * Can show that the second version of Euclid’s algorithm takes time proportional to the number of digits in m * If mis 1 billion (10%), the naive algorithm takes billions of steps, but this algorithm takes tens of stepsgra (10) 2) Efficiency bret lav dug galt.) >! * Can show that the second version of Euclid’s algorithm takes time proportional to the number of digits in m * Ifmis 1 billion (10%, the naive algorithm takes billions of steps, but this algonthm takes tens of stepsInstalling Python * Python is available on all platforms: Linux, MacOS and Windows : * Two main flavours of Python * Python 2.7 * Python 3+ (currently 3.5.x) * We will work with Python 3+Python 2.7 vs Python 3 * Python 2.7 is a “static” older version * Many libraries for scientific and statistical computing are still in Python 2.7, hence still “alive” * Python 3 is mostly identical to Python 2.7 * Designed to better incorporate new features * Will highlight some differences as we go alongDownloading Python 3.5 * Any Python 3 version should be fine, but the latest is 3.5.x * On Linux, it should normally be installed by default, else use the package manager * For MacOS and Windows, download and install from https://www.python.org/downloads/release/ python-350/ * If you have problems installing Python, search online or ask someone!Downloading Python 3.5 * Any Python 3 version should be fine, but the latest ts 3.5.x * On Linux, it should normally be installed by default, else use the package manager * For MacOS and Windows, download and install from https://www.python.org/downloads/release/ python-350/ * If you have problems installing Python, search online or ask someone!Interpreters vs compilers * Programming languages are “high level”, for humans to understand * Computers need “lower level” instructions * Compiler: Translates high level programming language to machine level instructions, generates “executable” code * Interpreter: Itself a program that runs and directly “understands” high level programming languageInterpreters vs compilers * Programming languages are “high level”, for humans to understand * Computers need “lower level” instructions * Compiler: Translates high level programming language to machine level instructions, generates “executable” code * Interpreter: Itself a program that runs and directly “understands” high level programming languagePython interpreter * Python is basically an interpreted language * Load the Python interpreter * Send Python commands to the interpreter to be executed * Easy to interactively explore language features * Can load complex programs from files * >>> from filename import *Installing Python * Python is available on all platforms: Linux, MacOS and Windows * Two main flavours of Python * Python 2.7 « Python 3+ (currently 3.5.x) * We will work with Python 3+Python 2.7 vs Python 3 * Python 2.7 is a “static” older version * Many libraries for scientific and statistical computing are still in Python 2.7, hence still “alive” * Python 3 is mostly identical to Python 2.7 * Designed to better incorporate new features * Will highlight some differences as we go alongDownloading Python 3.5 * Any Python 3 version should be fine, but the latest is 3.5.x * On Linux, it should normally be installed by default, else use the package manager * For MacOS and Windows, download and install from https://www.python.org/downloads/release/ python-350/ * If you have problems installing Python, search online or ask someone!Interpreters vs compilers * Programming languages are “high level”, for humans to understand * Computers need “lower level” instructions * Compiler: Translates high level programming lanquage to machine level instructions, generates “executable” code * Interpreter: Itself a program that runs and directly “understands” high level programming languagePython interpreter * Python is basically an interpreted language * Load the Python interpreter * Send Python commands to the interpreter to be executed * Easy to interactively explore language features * Can load complex programs from files * >>> from filename import *[ranecorn a} Lee Le ono 8) ac si) ae) fm-0 Le Ua SDD) acted] tase SP -0 for f in fo ay Oto) retern(cf[-1)>. tiene perenne cere ees Me serait ME en edad edone py eee ed peal sienelicl Aieetethd aed Soddn oh Ta LaD. A Aah e BEAR le trait ed @.2 (clang- pers pet oe aot) Pint Sieh ee ee! cense” for mere inforastionetd brSome resources * The online Python tutorial is a good place to start: https://docs.python.org/3/tutorial/index.ntm! * Here are some books, again available online: * Dive into Python 3, Mark Pilgrim http://www.diveintopython3.net/ * Think Python, 2nd Edition, Allen B. Downey http://greenteapress.com/wp/think-python-2e/Learning programming * Programming cannot be learnt theoretically * Must write and execute your code to fully appreciate the subject * Python syntax is light and is relatively easy to learn * Go for it!