summaryrefslogtreecommitdiffstats
path: root/dijo-git
diff options
context:
space:
mode:
Diffstat (limited to 'dijo-git')
0 files changed, 0 insertions, 0 deletions
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2026-04-24 23:21:45 +0000
'>73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Lachaise Assignment
% LaTeX Template
% Version 1.0 (26/6/2018)
%
% This template originates from:
% http://www.LaTeXTemplates.com
%
% Authors:
% Marion Lachaise & François Févotte
% Vel (vel@LaTeXTemplates.com)
%
% License:
% CC BY-NC-SA 3.0 (http://creativecommons.org/licenses/by-nc-sa/3.0/)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\documentclass{article}
\input{structure.tex}

\title{CENG567: Homework \#1}
\author{Yiğit Sever}
\date{\today}

%----------------------------------------------------------------------------------------

\begin{document}

\maketitle

\section{Stable Matching}%
\label{sec:stable_matching}

% question a ✅ {{{1 %

\begin{question}%
    \label{q:1_a}
    Use \emph{Gale-Shapley} algorithm to find a stable matching for the following set of four colleges, four students and their preference lists.
\end{question}

\begin{commandline}[Gale-Shapley algorithm, from lecture slides, edited for the context]
    \begin{verbatim}
    Initialize each person to be free.
    while (some student is free and hasn't applied to every college) {
        Choose such a student m
        c = 1st college on m's list to whom m has not yet applied
        if (c is free)
            assign c to m for potential application (a)
        else if (c prefers m to their current applicant m')
            assign m and c for potential application, and m' to be free (b)
        else
            c rejects m (c)
    }
    \end{verbatim}
 \end{commandline}

A quick trace of the algorithm;

\begin{enumerate}
    \item $S_1$ is free;
        \begin{enumerate}
            \item applies to first college on their preference list $C_4$;
            \item $C_4$ is free so it accepts and is matched with $S_1$ (a).
        \end{enumerate}
    \item $S_2$ is free
        \begin{enumerate}
            \item applies to first college on their preference list; $C_1$
            \item $C_1$ is free so it accepts and is matched with $S_2$ (a).
        \end{enumerate}
    \item $S_3$ is free;
        \begin{enumerate}
            \item applies to first college on their preference list; $C_1$
            \item $C_1$ rejects $S_3$ because it prefers $S_2$ to $S_3$ (c).
            \item applies to second college on their preference list; $C_2$
            \item $C_2$ is free so it accepts and is matched with $S_3$ (a).
        \end{enumerate}
    \item $S_4$ is free;
        \begin{enumerate}
            \item applies to first college on their preference list; $C_4$
            \item $C_4$ rejects $S_4$ because it prefers $S_1$ to $S_4$ (c).
            \item applies to second college on their preference list; $C_3$
            \item $C_3$ is free so it accepts and is matched with $S_4$ (a).
        \end{enumerate}
        \item There are no more free students to match, algorithm terminates.
\end{enumerate}

The final matching and the answer to Question~\ref{sec:stable_matching}(a) is;

\begin{align*}
    S_1 &\rightarrow C_4 \\
    S_2 &\rightarrow C_1 \\
    S_3 &\rightarrow C_2 \\
    S_4 &\rightarrow C_3
\end{align*}

% 1}}} %

% question b ✅ {{{1 %

\begin{question}
    Find another stable matching with the same algorithm.
\end{question}

All executions of \emph{Gale-Shapley} yield the same stable matching (that is proposer-optimal) and cannot produce \emph{another} stable matching like the question text asks for.

% 1}}} %

% question c ✅ {{{1 %

\begin{question}
    Consider a pair of man $m$ and woman $w$ where $m$ has $w$ at the top of his preference list and $w$
    has $m$ at the top of her preference list. Does it always have to be the case that the pairing $(m, w)$ exist
    in every possible stable matching? If it is true, give a short explanation. Otherwise, give a counterexample.
\end{question}

\emph{Proof by contradiction}.
Assume that in the resulting matching of \emph{Gale-Shapley}, we have $S'$, where $m$ is matched with $w'$ and $w$ is matched with $m'$.

The definition of stable matching dictates that there is \emph{no incentive to exchange}, yet in $S'$ $m$ can trade up to $w$ since $m$ prefers $w$ to $w'$ and $w$ can trade up since $w$ prefers $m$ to $m'$.

$S'$ could not have occurred since men propose in accordance to their preference list, which $w$ is on top of for $m$ and no other men that may propose to $w$ can make $w$ switch since they are not more preffered than $m$.

% 1}}} %

% question d ✅ {{{1 %

\begin{question}
    Give an instance of $n$ colleges, $n$ students, and their preference lists so that the Gale-Shapley algorithm requires only $O(n)$ iterations, and prove this fact.
\end{question}

For a proposer agnostic formation, arrange the preference lists of colleges and students as follows;

\begin{gather*}%
    \label{eq:student_prefs}
    C_{1} \rightarrow \left\{ S_{1}, S_{n}, S_{n-1}, \dots, S_{2} \right\} \\
    C_{2} \rightarrow \left\{ S_{2}, S_{n}, S_{n-1}, \dots, S_{3} \right\} \\
    \dots  \\
    C_{k} \rightarrow \left\{ S_{k}, S_{n}, S_{n-1}, \dots, S_{k+1} \right\} \\
    \dots \\
    C_{n} \rightarrow \left\{ S_{n}, S_{n-1}, S_{n-2}, \dots, S_{1} \right\} \\
\end{gather*}

\begin{gather*}%
    \label{eq:student_prefs}
    S_{1} \rightarrow \left\{ C_{1}, C_{n}, C_{n-1}, \dots, C_{2} \right\} \\
    S_{2} \rightarrow \left\{ C_{2}, C_{n}, C_{n-1}, \dots, C_{3} \right\} \\
    \dots  \\
    S_{k} \rightarrow \left\{ C_{k}, C_{n}, C_{n-1}, \dots, C_{k+1} \right\} \\
    \dots \\
    S_{n} \rightarrow \left\{ C_{n}, C_{n-1}, C_{n-2}, \dots, C_{1} \right\} \\
\end{gather*}

Where the student list $\{S_1, S_2, \dots, S_{n}\}$ and college list $\{C_1, C_2, \dots, C_{n}\}$ are shifted.

In this setup, ever proposer will propose to the first suitor in their preference list which is guaranteed to be free since they are not the first on any other suitor's preference list.

The algorithm in this instance runs in $\mathcal{O}(n)$ iterations, every proposer will follow the (a) branch in the algorithm given under Question~\ref{q:1_a}.

% 1}}} %

% question e ✅ {{{1 %

\begin{question}
    Give another instance for which the algorithm requires $\Omega(n^{2})$ iterations (that is, it requires at least $cn^{2}$ iterations for some constant $0 < c \le 1)$, and prove this fact.
\end{question}

\begin{infor}
    We have collaborated with CENG567 student Manolya Atalay for this question and used the answers and explanations given in the linked Mathematics Stack Exchange question\footnote{\url{https://math.stackexchange.com/questions/1410898/worst-case-for-the-stable-marriage-problem}}.
\end{infor}

The worst problem instance for the algorithm (the instance that requires the largest number of steps) can be inferred as follows;

The highest number of times a man $m$ can \emph{propose} (every iteration has one proposal in it) to $n$ many women is $n-1$. It cannot be $n$ because it would mean that $m$ did not find a suitable partner (rejected by everyone), which is contradictory with the algorithm's perfect matching guarantee.

For $n$ men, in the worst case, each one can propose $n-1$ times; $n(n-1)$. One man has to propose one last time after getting rejected by $n-1$ women, giving the total number of proposals;

\begin{equation}
    n (n-1) + 1
\end{equation}

As the theoretical highest number of iterations \emph{Gale-Shapley} can have.

As for the instance that produces this run time;

Men have to arrange their preference list as follows for $n>2$.

\begin{equation}
    M_{k} \rightarrow
    \begin{cases}
        W_{k}, W_{k+1}, \dots, W_{n-1}, W_{1}, W_{2}, \dots, W_{n} &\mbox{if } k \ne n \\
        W_{k-1}, W_{k}, \dots, W_{n-1}, W_{1}, W_{2}, \dots, W_{n} &\mbox{if } k = n \\
    \end{cases}
\end{equation}

In other words, the \emph{last} man's preference list is identical to one above them.

Women have to arrange their preference list as follows for $n>2$;

\begin{equation}
    W_{k} \rightarrow
    \begin{cases}
        M_{k+1}, M_{k}, \dots &\mbox{if } k \ne n-1 \\
        M_{1}, M_{n}, \dots &\mbox{if } k = n-1 \\
    \end{cases}
\end{equation}

The preference list of the last women $W_{n}$ does not matter neither does the order of the rest of the men in the preference list of the women given.

We have implemented \emph{Gale-Shapley} and an instance set creation script to test our findings;

\begin{commandline}[n=3]
    \begin{verbatim}
    n: Men Preference Table
    A | 1 2 3
    B | 2 1 3
    C | 2 1 3
    ----
    Women Preference Table
     1| B A C
     2| A C B
     3| A B C
    ----
    Matching B with 2 (a)
    Matching C with 2, B is now free (b)
    Matching A with 1 (a)
    Matching B with 1, A is now free (b)
    Matching A with 2, C is now free (b)
    C is rejected by 1 because B is more preffered (c)
    Matching C with 3 (a)
    Final pairings:
    A - 2
    B - 1
    C - 3

    A proposed 2 times
    B proposed 2 times
    C proposed 3 times
    total of 7 times with n(n-1) + 1 = 7
    \end{verbatim}
\end{commandline}

\begin{commandline}[n=6]
    \begin{verbatim}
    n: Men Preference Table
    A | 1 2 3 4 5 6
    B | 2 3 4 5 1 6
    C | 3 4 5 1 2 6
    D | 4 5 1 2 3 6
    E | 5 1 2 3 4 6
    F | 5 1 2 3 4 6
    ----
    Women Preference Table
     1| B A C D E F
     2| C B A D E F
     3| D C A B E F
     4| E D A B C F
     5| A F B C D E
     6| A B C D E F
    ----
    Matching D with 4 (a)
    Matching A with 1 (a)
    Matching F with 5 (a)
    E is rejected by 5 because F is more preffered (c)
    Matching C with 3 (a)
    Matching B with 2 (a)
    // --- edited out for space ---
    C is rejected by 1 because A is more preffered (c)
    Matching C with 2, B is now free (b)
    B is rejected by 3 because D is more preffered (c)
    B is rejected by 4 because E is more preffered (c)
    B is rejected by 5 because F is more preffered (c)
    Matching B with 1, A is now free (b)
    A is rejected by 2 because C is more preffered (c)
    A is rejected by 3 because D is more preffered (c)
    A is rejected by 4 because E is more preffered (c)
    Matching A with 5, F is now free (b)
    F is rejected by 1 because B is more preffered (c)
    F is rejected by 2 because C is more preffered (c)
    F is rejected by 3 because D is more preffered (c)
    F is rejected by 4 because E is more preffered (c)
    Matching F with 6 (a)
    Final pairings:
    E - 4
    B - 1
    A - 5
    D - 3
    C - 2
    F - 6

    A proposed 5 times
    B proposed 5 times
    C proposed 5 times
    D proposed 5 times
    E proposed 5 times
    F proposed 6 times
    total of 31 times with n(n-1) + 1 = 31
   \end{verbatim}
\end{commandline}



% 1}}} %

\section{Stable Matching Variation}%
\label{sec:stable_matching_variation}

% stable matching question ✅ {{{1 %

\begin{question}
    Consider a Stable Matching problem with men and women.
    Consider a woman $w$ where she prefers man $m$ to $m'$, but both $m$ and $m'$ are low on her list of preferences.
    Can it be the case that by switching the order of $m$ and $m'$ on her list of preferences (i.e., by falsely claiming that she prefers $m'$ to $m$) and running the algorithm with this modified preference list, $w$ will end up with a man $m''$ that she prefers to both $m$ and $m'$?
    Either give a proof that shows such an improvement is impossible, or give an example preference list for which an improvement for $w$ is possible.
\end{question}

We will give an example instance.
Let's take the example preference list for men; $M = {x,y,z}$ and women: $W = {a,b,c}$.

\begin{table}[!htb]
    \caption{Preference lists of women}%
    \label{table:2_general}
    \begin{minipage}{.5\linewidth}
        \caption{Everyone is being truthful}%
        \label{table:2_true}
        \centering
        \begin{tabular}{l|lll}
            \textbf{a} & y & x & z \\
            \textbf{b} & x & y & z \\
            \textbf{c} & x & y & z
        \end{tabular}
    \end{minipage}%
    \begin{minipage}{.5\linewidth}
        \centering
        \caption{a is lying about their preference}%
        \label{table:2_lie}
        \begin{tabular}{l|lll}
            \textbf{a} & y & z & x \\
            \textbf{b} & x & y & z \\
            \textbf{c} & x & y & z
        \end{tabular}
        \begin{tikzpicture}[overlay]
            \draw[red, line width=1pt] (-0.72,0.52) ellipse (0.70cm and 0.2cm);
        \end{tikzpicture}
    \end{minipage}
\end{table}

In the example given above in Table~\ref{table:2_general}, we have two cases for women, Table~\ref{table:2_true} where everyone has given their actual preference list and Table~\ref{table:2_lie} where $a$ has lied about their \nth{2} and \nth{3} preferences.

The following is the men's preference table;

\begin{table}[htb]
    \centering
    \begin{tabular}{l|lll}
        \textbf{x} & a & b & c \\
        \textbf{y} & b & a & c \\
        \textbf{z} & a & b & c
    \end{tabular}
    \caption{The preference lists of the men.}%
    \label{tab:men_pref}
\end{table}

The trace of \emph{Gale-Shapley} algorithm for the truth telling case with Table~\ref{table:2_true} and Table~\ref{tab:men_pref} is below;

\begin{enumerate}
    \item Matching x with a (a)
    \item Matching y with b (a)
    \item z proposes to a but gets rejected because a prefers x more (c)
    \item z proposes to b but gets rejected because b prefers y more (c)
    \item Matching z with c (a)
\end{enumerate}

Which produces the following matching. Note that $a$ is matched with their \nth{2} choice.

\begin{align*}
    x &\rightarrow a \\
    y &\rightarrow b \\
    z &\rightarrow c
\end{align*}

Now let's examine the trace of the algorithm when $a$ lies by providing an altered preference table. The trace below uses Table~\ref{table:2_lie} and Table~\ref{tab:men_pref};

\begin{enumerate}
    \item Matching y with b (a)
    \item Matching z with a (a)
    \item \textbf{x proposes to a but gets rejected because a lies by saying that they like z more (c)}
    \item Matching x with b, y is now free (b)
    \item \textbf{Matching y with a, z is now free (b)}
    \item z proposes to b but gets rejected because b prefers x more (c)
    \item Matching z with c (a)
\end{enumerate}

The highlighted lines show that $a$ lies when proposed by $x$ and then is able to trade up to $y$. The following it the final matching;

\begin{align*}
    x &\rightarrow b \\
    y &\rightarrow a \\
    z &\rightarrow c
\end{align*}

Woman $a$ was able to get their highest preferred option $y$ by telling a lie. We have proved that an improvement is possible.

% 1}}} %

\section{Asymptotics}%
\label{sec:asymptotics}

% asymptotics ✅ {{{1 %

\begin{question}
    What is the running time of this algorithm as a function of $n$? Specify a function $f$ such that the running time of the algorithm is $\Theta(f(n))$.
\end{question}

{\centering
    \begin{minipage}{.7\linewidth}
        \begin{algorithm}[H]
            \For(\tcc*[f]{outer loop}){$i = 2$; $i < n$; $i += 1$ }{
                \For(\tcc*[f]{inner loop}){$j=1$; $j < n$; $j * = i$}{
                    Some $\Theta(1)$ operation\;
                }
            }
            \caption{Equivalent algorithm to one given in Question 3, edited for brevity}%
            \label{alg:question_3}
\end{algorithm}
\end{minipage}
\par
}

The outer loop runs in linear time $\mathcal{O}(n)$ whereas the inner loop requires some deconstruction;

Take the first iteration of the inner loop, $i = 2$ and $j$ is initialized at $1$.

\begin{align*}%
    \label{eq:3_iterations}
    1^{\text{st}} ~ \text{iteration} &\rightarrow  j = j * i \implies  j = 2 \\
    2^{\text{nd}} ~ \text{iteration} &\rightarrow  j = 4 \\
    3^{\text{rd}} ~ \text{iteration} &\rightarrow  j = 8 \\
    \dots \\
    m^{\text{th}} ~ \text{iteration} &\rightarrow  j = i^{m} < n \\
\end{align*}

$m$ is the last iteration of the inner loop because it hit the stopping condition $i^{m} < n$.
Using the Equations above we can find the running time for the inner loop to hit stopping condition;

\begin{gather*}
    i^{m} < n \\
    m < \log_{i} n
\end{gather*}

In other words, the inner loop has a running time in the order of $\mathcal{O}(\log{n})$. By the product property of $\mathcal{O}$-notation we have the running time of $\mathcal{O}(n\log_{a}{n}), a > 1$ for the entire algorithm. The base $a$ is useful to answer the rest of the question;

We can specify a function $f$ such as $f = n \log_{b}(n)$ where $b > 1$. From the course slides;

\begin{equation*}
    \lim_{n \to \infty}\frac{n\log_{a}{n}}{n \log_{b}{n}} = c
\end{equation*}

And when the limit of two functions $f, g$ converge to some constant $c$, then $f(n) = \Theta(g(n))$.

% 1}}} %

\section{Big $\mathcal{O}$ and $\Omega$}%
\label{sec:big_o_and_omega_}

% question a ✅ {{{1 %

\begin{question}
    Let $f(n)$ and $g(n)$ be asymptotically positive functions. Prove or disprove the following conjectures.
\end{question}

\begin{info}[]
    $f(n) = \mathcal{O}(g(n))$ implies $g(n) = \mathcal{O}(f(n))$
\end{info}

From the definition of Big $\mathcal{O}$ notation given in the lecture slides;

\begin{equation}%
    \label{eq:a_bigo}
    \exists ~ c > 0 \quad \text{and} \quad n_{0} \ge 0 \quad \mid \quad 0 \le f(n) \le c \cdot g(n) \quad \forall n \ge n_{0}
\end{equation}

If we rearrange the right hand side of the Proposition~\ref{eq:a_bigo};

\begin{equation*}%
    0 \le \frac{1}{c} f(n) \le g(n) \\
\end{equation*}

Or simply,

\begin{equation}
    \label{eq:a_bio_rearranged}
    0 \le c' f(n) \le g(n)
\end{equation}

By the definition of Big $\mathcal{O}$ notation, in order for $g(n) = \mathcal{O}(f(n))$ to be true,

\begin{equation*}
    \exists ~ k > 0 \quad \text{and} \quad n_{0}' \ge 0 \quad \mid \quad 0 \le g(n) \le k \cdot f(n) \quad \forall n' \ge n_{0}'
\end{equation*}

Has to be true, yet from Equation~\ref{eq:a_bio_rearranged} we know that $f(n)$ multiplied by some constant $c'$ is strictly smaller than $g(n)$. The conjecture is \emph{false}.

\begin{info}[]
    $f(n) = \mathcal{O}((f(n))^{2})$
\end{info}

% https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-046j-introduction-to-algorithms-sma-5503-fall-2005/assignments/ps1sol.pdf

We can reuse the definition of Big $\mathcal{O}$ notation given in Equation~\ref{eq:a_bigo}, and using the question text, we have;

For $c > 1$ and $n_{0} > 1$;

\begin{align*}%
    f(n) &\le f(n)^{2} \quad \forall n \ge n_{0} \\
    \text{take} ~ g(n) &= f(n)^{2} \\
    f(n) &\le c\cdot g(n) \quad \forall n \ge n_{0}
\end{align*}

For $f(n) \ge 1$. The conjecture is true for asymptotically positive function $f(n)$ but does not hold for an $f(n) < 1$.

\begin{info}[]
    $f(n) + o(f(n)) = \Theta(f(n))$
\end{info}

From the definition of Big Theta notation, we are trying to prove a relation such that;

\begin{equation}%
    \label{eq:a_big_theta}
    0 \le c_{1} \cdot f(n) \le f(n) + o(f(n)) \le c_{2} \cdot f(n)
\end{equation}

First, let's give the little-o notation to remove $o(f(n))$ from Equation~\ref{eq:a_big_theta};

\begin{equation}%
    \label{eq:a_little_o}
    \forall ~ c > 0 \quad \exists ~ n_{0} > 0 \mid ~ 0 \le g(n) < c \cdot f(n) ~ \forall n \ge n_{0}
\end{equation}

So we can rewrite Equation~\ref{eq:a_big_theta} using Equation~\ref{eq:a_little_o};

\begin{equation}%
    \label{eq:a_rewritten}
    c_1 \cdot f(n) \le f(n) + g(n) \le c_2 \cdot f(n)
\end{equation}

It's trivial to pick $c_1 < 1$ to deal with the left hand side of the inequality;

\begin{equation*}
    f(n) \le f(n) + g(n)
\end{equation*}

For the right hand side; we can start by picking $n_{0} = 1$ and $c = 1$ in the Equation~\ref{eq:a_little_o}; keeping $g(n)$ strictly smaller than  $f(n)$.

In order to prove the right hand side of the inequality; we can pick $c_2 > 2$ which equals;

\begin{eqnarray*}%
    \label{eq:c2