# Efficient computation

- Describe at a high level some interesting computational problems.
- The difference between polynomial and exponential time.
- Examples of techniques for obtaining efficient algorithms
- Examples of how seemingly small differences in problems can make (at least apparent) huge differences in their computational complexity.

“The problem of distinguishing prime numbers from composite and of resolving the latter into their prime factors is … one of the most important and useful in arithmetic … Nevertheless we must confess that all methods … are either restricted to very special cases or are so laborious … they try the patience of even the practiced calculator … and do not apply at all to larger numbers.”, Carl Friedrich Gauss, 1798

“For practical purposes, the difference between algebraic and exponential order is often more crucial than the difference between finite and non-finite.”, Jack Edmunds, “Paths, Trees, and Flowers”, 1963

“Sad to say, but it will be many more years, if ever before we really understand the Mystical Power of Twoness… 2-SAT is easy, 3-SAT is hard, 2-dimensional matching is easy, 3-dimensional matching is hard. Why? oh, Why?”Eugene Lawler

So far we were concerned with which functions are computable and which
ones are not. But now we return to *quantitative considerations* and
study the time that it takes to compute functions mapping strings to
strings, as a function of the input length. One of the interesting
phenomenona of computing is that there is often a kind of a “thershold
phenomenon” or “zero one law”
for running time, where for many natural problems, they either can be
solved in *polynomial* running time (e.s., something like \(O(n^2)\) or
\(O(n^3)\)), or require *exponential* (e.g., at least \(2^{\Omega(n)}\) or
\(2^{\Omega(\sqrt{n})}\)) running time. The reasons for this phenomenon
are still not fully understood, but some light on this is shed by the
concept of *NP completeness*, which we will encounter later.

In this lecture we will survey some examples of computational problems, for some of which we know efficient (e.g., \(n^c\)-time for a small constant \(c\)) algorithms, and for others the best known algorithms are exponential. We want to get a feel as to the kinds of problems that lie on each side of this divide and also see how some seemingly minor changes in formulation can make the (known) complexity of a problem “jump” from polynomial to exponential.

In this lecture, we will not formally define the notion of running time, and so use the same notion of an \(O(n)\) or \(O(n^2)\) time algorithms as the one you’ve seen in an intro to CS course: “I know it when I see it”. In the next lecture, we will define this notion precisely, using our NAND++ and NAND<< programming languages. One of the nice things about the theory of computation is that it turns out that, like in the context of computability, the details of the precise computational model or programming language don’t matter that much, especially if you mostly care about the distinction between polynomial and exponential time.

## Problems on graphs

We nor present a few examples of computational problems that people are
interesting in solving. Many of the problems will involve *graphs*. We
have already encountered graphs in the context of Boolean circuits, but
let us now quickly recall the basic notation. A graph \(G\) consists of a
set of *vertices* \(V\) and *edges* \(E\) where each edge is a pair of
vertices. In a *directed* graph, an edge is an ordered pair \((u,v)\),
which we sometimes denote as \(\overrightarrow{u\;v}\). In an *undirected*
graph, an edge is an unordered pair (or simply a set) \(\{ u,v \}\) which
we sometimes denote as \(\overline{u\; v}\) or \(u \sim v\).An equivalent viewpoint is that an undirected graph is like a
directed graph with the property that whenever the edge
\(\overrightarrow{u\; v}\) is present then so is the edge
\(\overrightarrow{v\; u}\). We will
assume graphs are undirected and *simple* (i.e., containing no parallel
edges or self-loops) unless stated otherwise.

We typically will think of the vertices in a graph as simply the set
\([n]\) of the numbers from \(0\) till \(n-1\). Graphs can be represented
either in the *adjacency list* representation, which is a list of \(n\)
lists, with the \(i^{th}\) list corresponding to the neighbors of the
\(i^{th}\) vertex, or the *adjacency matrix* representation, which is an
\(n\times n\) matrix \(A\) with \(A_{i,j}\) equalling \(1\) if the edge
\(\overrightarrow{u\; v}\) is present and equalling \(0\) otherwise.In an undirected graph, the adjacency matrix \(A\) is *symmetric*,
in the sense that it satisfies \(A_{i,j}=A_{j,i}\). We
can transform between these two representations using \(O(n^2)\)
operations, and hence for our purposes we will mostly consider them as
equivalent. We will sometimes consider *labeled* or *weighted* graphs,
where we assign a label or a number to the edges or vertices of the
graph, but mostly we will try to keep things simple and stick to the
basic notion of an unlabeled, unweighted, simple undirected graph.

There is a reason that graphs are so ubiquitous in computer science and other sciences. They can be used to model a great many of the data that we encounter. These are not just the “obvious” networks such as the road network (which can be thought of as a graph of whose vertices are locations with edges corresponding to road segments), or the web (which can be thought of as a graph whose vertices are web pages with edges corresponding to links), or social networks (which can be thought of as a graph whose vertices are people and the edges correspond to friend relation). Graphs can also denote correlations in data (e.g., graph of observations of features with edges corresponding to features that tend to appear together), casual relations (e.g., gene regulatory networks, where a gene is connected to gene products it derives), or the state space of a system (e.g., graph of configurations of a physical system, with edges corresponding to states that can be reached from one another in one step).

We now give some examples of computational problems on graphs. As mentioned above, to keep things simple, we will restrict attention to undirected simple graphs. In all cases the input graph \(G=(V,E)\) will have \(n\) vertices and \(m\) edges.

### Finding the shortest path in a graph

The *shortest path problem* is the task of, given a graph \(G=(V,E)\) and
two vertices \(s,t \in V\), to find the length of the shortest path
between \(s\) and \(t\) (if such a path exists). That is, we want to find
the smallest number \(k\) such that there are vertices
\(v_0,v_1,\ldots,v_k\) with \(v_0=s\), \(v_k=t\) and for every
\(i\in\{0,\ldots,k-1\}\) an edge between \(v_i\) and \(v_{i+1}\). If each
vertex has at least two neighbors then there can be an *exponential*
number of paths from \(s\) to \(t\), but fortunately we do not have to
enumerate them all to find the shortest path. We can do so by performing
a breadth first search
(BFS), enumerating
\(s\)’s neighbors, and then neighbors’ neighbors, etc.. in order. If we
maintain the neighbors in a list we can perform a BFS in \(O(n^2)\) time,
while using a queue we can do this in \(O(m)\) time.A *queue* stores a list of elements in “First In First Out (FIFO)”
order and so each “pop” operation removes an element from the queue
in the order that they were “pushed” into it; see the Wikipedia
page. Since we assume \(m \geq n-1\), \(O(m)\)
is the same as \(O(n+m)\). Dijkstra’s
algorithm is a well-known generalization of
BFS to *weighted* graphs.

More formally, the algorithm for shortest path can be described as follows:

**Algorithm \(SHORTESTPATH\):**

**Input:**Graph \(G=(V,E)\), vertices \(s,t\)**Goal:**Find the shortest path \(v_0,v_1,\ldots,v_k\) such that \(v_0=s\), \(v_k=t\) and \(\{ v_i,v_{i+1} \} \in E\) for every \(i\in [k]\), if such a path exists.**Operation:**- We will maintain a label \(L[v]\) for every vertex \(v\). Initially
no vertex is labeled except for \(s\) that is labeled with
“start”.
- We maintain a
*queue*\(Q\) of vertices, initially \(Q\) contains only \(s\). - While \(Q\) is not empty do the following:

a. Pop the vertex \(v\) from the top of the queue.

b. If \(v=t\) exit output the path which is the reverse order of \(v,L[v],L[L[v]],L[L[L[v]]],\ldots,s\).

c. Otherwise, label all the unlabeled neighbors of \(v\) with \(v\) and add them to \(Q\) - Output “no path”

- We will maintain a label \(L[v]\) for every vertex \(v\). Initially
no vertex is labeled except for \(s\) that is labeled with
“start”.

Since we only add to the queue unlabeled vertices, we never push to the queue a vertex more than once, and hence the algorithm takes \(n\) “push” and “pop” operations. It returns the correct answer since add the vertices to the queue in the order of their distance from \(s\), and hence we will reach \(t\) after we have explored all the vertices that are closer to \(s\) than \(t\).

### Finding the longest path in a graph

The *longest path problem* is the task of, given a graph \(G=(V,E)\) and
two vertices \(s,t \in V\), to find the length of the *longest* simple
(i.e., non intersecting) path between \(s\) and \(t\). If the graph is a
road network, then the longest path might seem less motivated than the
shortest path, but of course graphs can be and are used to model a
variety of phenomena, and in many such cases the longest path (and some
of its variants) are highly moticated. In particular, finding the
longest path is a generalization of the famous Hamiltonian path
problem which
asks for a *maximally long* simple path (i.e., path that visits all \(n\)
vertices once) between \(s\) and \(t\), as well as the notorious traveling
salesman problem
(TSP) of
finding (in a weighted graph) a path visiting all vertices of cost at
most \(w\). TSP is a classical optimization problem, with applications
ranging from planning and logistics to DNA sequencing and astronomy.

A priori it is not clear that finding the longest path should be harder than finding the shortest path, but this turns out to be the case. While we know how to find the shortest path in \(O(n)\) time, for the longest path problem we have not been able to significantly improve upon the trivial brute force algorithm that tries all paths.

Specifically, in a graph of degree at most \(d\), we can enumerate over all paths of length \(k\) by going over the (at most \(d\)) neighbors of each vertex. This would take about \(O(d^k)\) steps, and since the longest simple path can’t have length more than the number of vertices, this means that the brute force algorithms runs in \(O(d^n)\) time (which we can bound by \(O(n^n)\) since the maximum degree is \(n\)). The best algorithm for the longest path improves on this, but not by much: it takes \(\Omega(c^n)\) time for some constant \(c>1\).At the moment the best record is \(c \sim 1.65\) or so. Even obtaining an \(O(2^n)\) time bound is not that simple, see Reference:longest-path-ex.

### Finding the minimum cut in a graph

Given a graph \(G=(V,E)\), a *cut* is a subset \(S\) of \(V\) such that \(S\) is
neither empty nor is it all of \(V\). The edges cut by \(S\) are those edges
where one of their endpoints is in \(S\) and the other is in
\(\overline{S} = V \setminus S\). We denote this set of edges by
\(E(S,\overline{S})\). If \(s,t \in V\) then an *\(s,t\) cut* is a cut such
that \(s\in S\) and \(t\in \overline{S}\). (See Reference:cutingraphfig.)
The *minimum \(s,t\) cut problem* is the task of finding, given \(s\) and
\(t\), the minimum number \(k\) such that there is an \(s,t\) cut cutting \(k\)
edges (the problem is also sometimes phrased as finding the set that
achieves this minimum).One can also define the problem of finding the *global minimum
cut* (i.e., the non-empty and non-everything set \(S\) that minimizes
the number of edges cut). One can verify that a polynomial time
algorithm for the minimum \(s,t\) cut can be used to solve the global
cut in polynomial time as well (can you see why?).

The minimum \(s,t\) cut problem appears in many applications. Minimum cuts
often correspond to *bottlenecks*. For example, in a communication
network the minimum cut between \(s\) and \(t\) corresponds to the smallest
number of edges that, if dropped, will disconnect \(s\) from \(t\). Similar
applications arise in scheduling and planning. In the setting of image
segmentation, one can
define a graph whose vertices are pixels and whose edges correspond to
neighboring pixels of distinct colors. If we want to separate the
foreground from the background then we can pick (or guess) a foreground
pixel \(s\) and background pixel \(t\) and ask for a minimum cut between
them.

**Solving the minimum cut problem:** Here is an algorithm to solve the
minimum cut problem:

**Algorithm MINCUTNAIVE:**

**Input:**Graph \(G=(V,E)\) and two distinct vertices \(s,t \in V\)**Goal:**Return \(S\) s.t. \(s\in S\) and \(t\not\in S\) that minimizes \(|E(S,\overline{S})|\).**Operation:**- If \(V=\{s,t\}\) then return \(\{s \}\).
- Otherwise choose \(v\in V \setminus \{s,t\}\), and define \(G',G''\)
to be two graphs with vertex set \(V\setminus \{v \}\), where in
\(G'\) the edge set \(E'\) is obtained by making all of \(v\)’s
neighbors be neighbors of \(s\), and in \(G''\) the edge set \(E''\)
is obtained by making all of \(v\)’s neighbors be neighbors of
\(t\). That is, \(E'\) has the same edges as \(E\) except that in
edges involving \(v\) we replace \(v\) with \(s\), and \(E''\) has the
same edges as \(E\) except that in edges involving \(v\) we replace
\(v\) with \(t\).
- Compute recursively \(S'=MINCUTNAIVE(G',s,t)\) and
\(S''=MINCUTNAIVE(G'',s,t)\).
- If \(S'\cup \{v\}\) cuts fewer edges in \(G\) than \(S''\) then return \(S' \cup \{ v \}\). Otherwise return \(S''\).

- If \(V=\{s,t\}\) then return \(\{s \}\).

It is an excellent exercise for you to pause at this point and verify:
**(i)** that you understand what this algorithm does, **(2)** that you
understand why this algorithm will in fact return the minimum cut in the
graph, and **(3)** that you can analyze the running time of this
algorithm.

We can prove by induction that Algorithm \(MINCUTNAIVE\) does indeed return the minimum cut. Indeed, it definitely does so for graphs of two vertices. Now we assume by induction that \(MINCUTNAIVE\) solves the minimum cut problem for graphs of at most \(n-1\) vertices, and we will prove that it does so for graphs of \(n\) vertices. Indeed, under the inductive hypothesis, our recursive calls in step 3 return the minimum cuts \(S'\) and \(S''\) of the \(n-1\) vertex graphs \(G'\) and \(G''\) respectively. But if \(v\) is the vertex we choose in step 2, we can think of \(G'\) as simply a graph where we “merged” \(s\) and \(v\) to be a single vertex with the neighbors of both \(s\) and \(v\), and \(G''\) as the graph where we merged \(t\) and \(v\). So the \(s,t\) cuts in \(G'\) correspond to \(s,t\) cuts in \(G\) that don’t separate \(s\) and \(v\), while \(s,t\) cuts in \(G''\) correspond to \(s,t\) cuts in \(G\) that don’t separate \(t\) and \(v\). Since in the minimum cut \(S\), either \(s\) or \(t\) will be in the same side as \(v\), the best one out of the minimal cuts from \(G'\) and \(G''\) will be the minimum cut in \(G\).

The running time \(T(n)\) of \(MINCUTNAIVE\) on \(n\) vertex graphs can be described by the recursive equation \(T(n)=2T(n-1)+f(n)\) where \(f(n)\) is the time to execute the non recursive steps 1,2 and 4. In an algorithms course we might want to worry about the exact data structures we use to implement these steps, but for this course it is enough that we can do these steps in polynomial time (which is not hard to see). Nevertheless, this running time bound is terrible: even if \(f(n)\) was equal to \(1\) we would still get that \(T(n) \geq 2^n\)! Indeed, this recursive algorithm is nothing but a fancy description of the trivial algorithm that enumerates over all the roughly \(2^n\) (in fact, precisely \(2^{n-2}\)) sets \(S\) that contain \(s\) and don’t contain \(t\).

Since minimum cut is a problem we want to solve, this seems like bad
news. Luckily however we are able to find much faster algorithms that
run in *polynomial time* (which, as mentioned in the mathematical
background lecture, we denote by \(poly(n)\)) for this problem. There are
several algorithms to do so, but many of them rely on the Max Flow Min
Cut Theorem
that says that the minimum cut between \(s\) and \(t\) equals the maximum
amount of *flow* we can send from \(s\) to \(t\), if every edge has unit
capacity.A *flow* of capacity \(c\) from a *source* \(s\) to a *sink* \(t\) in a
graph can be thought of as describing how one would send some
quantity from \(s\) to \(t\) in the graph (e.g., sending \(c\) liters of
water (or any other matter one can partition arbitrarily), on pipes
described by the edges). Mathematically, a flow is captured by
assigning numbers to edges, and requiring that on any vertex apart
from \(s\) and \(t\), the amount flowing it is equal to the amount
flowing out, while in \(s\) there are \(c\) units flowing out and in \(t\)
there are \(c\) units flowing in. For example, this directly implies that the value of the
minimum cut problem is the solution for the following linear
program:A *linear program* is the task of maximizing or minimizing a
linear function of \(n\) real variables \(x_0,\ldots,x_{n-1}\) subject
to certain linear equalities and inequalities on the variables.

\[ \begin{split} \max_{x\in \R^m} F_s(x)-F_t(x) \text{s.t.}\\ \forall_{u \not\in \{s,t\}} F_u(x)=0 \end{split} \] where for every vertex \(u\) and \(x\in \R^m\), \(F_u(x) = \sum_{e \in E \text{s.t.} u \in e} x_e\).

Since there is a polynomial-time algorithm for linear programming, the minimum cut (or, equivalently, maximum flow) problem can be solved in polynomial time. In fact, there are much better algorithms for this problem, with currently the record standing at \(O(\min\{ m^{10/7}, m\sqrt{n}\})\).

### Finding the maximum cut in a graph

We can also define the *maximum cut* problem of finding, given a graph
\(G=(V,E)\) the subset \(S\subseteq V\) that *maximizes* the number of edges
cut by \(S\).We can also consider the variant where one is given \(s,t\) and
looks for the \(s,t\)-cut that maximizes the number of edges cut. The
two variants are equivalent up to \(O(n^2)\) factors in the running
time, but the we use the global max cut forumlation since it is more
common in the literature. Like its cousin the minimum cut problem, the maximum cut
problem is also very well motivated. For example, it arises in VLSI
design, and also has some surprising relation to analyzing the Ising
model in statistical
physics.

Once again, a priori it might not be clear that the maximum cut problem should be harder than minimum cut but this turns out to be the case. We do not know of an algorithm that solves this problem much faster than the trivial “brute force” algorithm that tries all \(2^n\) possibilities for the set \(S\).

### A note on convexity

There is an underlying reason for the sometimes radical difference
between the difficulty of maximizing and minimizing a function over a
domain. If \(D \subseteq \R^n\), then a function \(f:D \rightarrow R\) is
*convex* if for every \(x,y \in D\) and \(p\in [0,1]\)
\(f(px+(1-p)y) \leq pf(x) + (1-p)f(y)\). That is, \(f\) applied to the
\(p\)-weighted midpoint between \(x\) and \(y\) is smaller than the
\(p\)-weighted average value of \(f\). If \(D\) itself is convex (which means
that if \(x,y\) are in \(D\) then so is the line segment between them), then
this means that if \(x\) is a *local minimum* of \(f\) then it is also a
*global minimum*. The reason is that if \(f(y)<f(x)\) then every point
\(z=px+(1-p)y\) on the line segment between \(x\) and \(y\) will satisfy
\(f(z) \leq p f(x) + (1-p)f(y) < f(x)\) and hence in particular \(x\) cannot
be a local minimum. Intuitively, local minima of functions are much
easier to find than global ones: after all, any “local search” algorithm
that keeps finding a nearby point on which the value is lower, will
eventually arrive at a local minima.One example of such a local search algorithm is gradient
descent which takes
a small step in the direction that would reduce the value by the
most amount based on the current derivative. There are also
algorithms that take advantage of the *second derivative* (hence are
known as *second order methods*) to potentially converge faster. Indeed, under certain
technical conditions, we can often efficiently find the minimum of
convex functions, and this underlies the reason problems such as minimum
cut and shortest path are easy to solve. On the other hand, *maximizing*
a convex function (or equivalently, minimizing a *concave* function) can
often be a hard computational task. A *linear* function is both convex
and concave, which is the reason both the maximization and minimization
problems for linear functions can be done efficiently.

The minimum cut problem is not a priori a convex minimization task,
because the set of potential cuts is *discrete*. However, it turns out
that we can embed it in a continuous and convex set via the (linear)
maximum flow problem. The “max flow min cut” theorem ensuring that this
embedding is “tight” in the sense that the minimum “fractional cut” that
we obtain through the maximum-flow linear program will be the same as
the true minimum cut. Unfortunately, we don’t know of such a tight
embedding in the setting of the *maximum* cut problem.

The issue of convexity arises time and again in the context of
computation. For example, one of the basic tasks in machine learning is
*empirical risk minimization*. That is, given a set of labeled examples
\((x_1,y_1),\ldots,(x_m,y_m)\), where each \(x_i \in \{0,1\}^n\) and
\(y_i \in \{0,1\}\), we want to find the function
\(h:\{0,1\}^n \rightarrow \{0,1\}\) from some class \(H\) that minimizes the
*error* in the sense of minimizing the number of \(i\)’s such that
\(h(x_i) \neq y_i\). Like in the minimum cut problem, to make this a
better behaved computational problem, we often embed it in a continuous
domain, including functions that could output a real number and
replacing the condition \(h(x_i) \neq y_i\) with minimizing some
continuous *loss function* \(\ell(h(x_i),y_i)\).We also sometimes replace or enhance the condition that \(h\) is in
the class \(H\) by adding a *regularizing term* of the form \(R(h)\) to
the minimization problem, where \(R:H \rightarrow \R\) is some measure
of the “complexity” of \(h\). As a general rule, the larger or more
“complex” functions \(h\) we allow, the easier it is to fit the data,
but the more danger we have of “overfitting”. When this embedding
is *convex* then we are guaranteed that the global minimizer is unique
and can be found in polynomial time. When the embedding is *non convex*,
we have no such guarantee and in general there can be many global or
local minima. That said, even if we don’t find the global (or even a
local) minima, this continuous embedding can still help us. In
particular, when running a local improvement algorithm such as Gradient
Descent, we might still find a function \(h\) that is “useful” in the
sense of having a small error on future examples from the same
distribution.In machine learning parlance, this task is known as *supervised
learning*. The set of examples \((x_1,y_1),\ldots,(x_m,y_m)\) is known
as the *training set*, and the error on additional samples from the
same distribution is known as the *generalization error*, and can be
measured by checking \(h\) against a *test set* that was not used in
training it.

## Beyond graphs

Not all computational problems arise from graphs. We now list some other examples of computational problems that of great interest.

### The 2SAT problem

A *propositional formula* \(\varphi\) involves \(n\) variables
\(x_1,\ldots,x_n\) and the logical operators AND (\(\wedge\)), OR (\(\vee\)),
and NOT (\(\neg\), also denoted as \(\overline{\cdot}\)). We say that such a
formula is in *conjunctive normal form* (CNF for short) if it is an AND
of ORs of variables or their negations (we call a term of the form \(x_i\)
or \(\overline{x}_i\) a *literal*). For example, this is a CNF formula \[
(x_7 \vee \overline{x}_{22} \vee x_{15} ) \wedge (x_{37} \vee x_{22}) \wedge (x_{55} \vee \overline{x}_7)
\]

We say that a formula is a \(k\)-CNF it is an AND of ORs where each OR
involves exactly \(k\) literals. The 2SAT problem is to find out, given a
\(2\)-CNF formula \(\varphi\), whether there is an assignment
\(x\in \{0,1\}^n\) that *satisfies* \(\varphi\), in the sense that it makes
it evaluate to \(1\) or “True”.

Determining the satisfiability of Boolean formulas arises in many applications and in particular in software and hardware verification, as well as scheduling problems. The trivial, brute-force, algorithm for 2SAT will enumerate all the \(2^n\) assignments \(x\in \{0,1\}^n\) but fortunately we can do much better.

The key is that we can think of every constraint of the form
\(\ell_i \vee \ell_j\) (where \(\ell_i,\ell_j\) are *literals*,
corresponding to variables or their negations) as an *implication*
\(\overline{\ell}_i \Rightarrow \ell_j\), since it corresponds to the
constraints that if the literal \(\ell'_i = \overline{\ell}_i\) is true
then it must be the case that \(\ell_j\) is true as well. Hence we can
think of \(\varphi\) as a directed graph between the \(2n\) literals, with
an edge from \(\ell_i\) to \(\ell_j\) corresponding to an implication from
the former to the latter. It can be shown that \(\varphi\) is
unsatisfiable if and only if there is a variable \(x_i\) such that there
is a directed path from \(x_i\) to \(\overline{x}_i\) as well as a directed
path from \(\overline{x}_i\) to \(x_i\) (see Reference:twosat_ex). This
reduces 2SAT to the (efficiently solvable) problem of determining
connectivity in directed graphs.

### The 3SAT problem

The 3SAT problem is the task of determining satisfiability for 3CNFs. One might think that changing from two to three would not make that much of a difference for complexity. One would be wrong. Despite much effort, do not know of a significantly better than brute force algorithm for 3SAT (the best known algorithms take roughy \(1.3^n\) steps).

Interestingly, a similar issue arises time and again in computation, where the difference between two and three often corresponds to the difference between tractable and intractable. As Lawler’s quote alludes to, we do not fully understand the reasons for this phenomenon, though the notions of \(\mathbf{NP}\) completeness we will see later does offer a partial explanation. It may be related to the fact that optimzing a polynomial often amounts to equations on its derivative. The derivative of a a quadratic polynomial is linear, while the derivative of a cubic is quadratic, and, as we will see, the difference between solving linear and quadratic equations can be quite profound.

### Solving linear equations

One of the most useful problems that people have been solving time and again is solving \(n\) linear equations in \(n\) variables. That is, solve equations of the form

\[ \begin{split} a_{0,0}x_0 + a_{0,1}x_1 + \cdots + a_{0,{n-1}}x_{n-1} &= b_0 \\ a_{1,0}x_0 + a_{1,1}x_1 + \cdots + a_{1,{n-1}}x_{n-1} &= b_1 \\ \vdots + \vdots + \vdots + \vdots &= \vdots \\ a_{n-1,0}x_0 + a_{n-1,1}x_1 + \cdots + a_{n-1,{n-1}}x_{n-1} &= b_{n-1} \end{split} \]

where \(\{ a_{i,j} \}_{i,j \in [n]}\) and \(\{ b_i \}_{i\in [n]}\) are real (or rational) numbers. More compactly, we can write this as the equations \(Ax = b\) where \(A\) is an \(n\times n\) matrix, and we think of \(x,b\) are column vectors in \(\R^n\).

The standard Gaussian
elimination
algorithm can be used to solve such equations in polynomial time (i.e.,
determine if they have a solution, and if so, to find it).To analyze this fully we need to ensure that the bit complexity
of the numbers involved does not grow too much, but fortunately we
can indeed ensure this using Cramer’s
rule. Also, as is
usually the case when talking about real numbers, we do not care
much for the distinction between solving equations exactly and
solving them to arbitrarily good precision. As we
discussed above, if we are willing to allow some loss in precision, we
even have algorithms that handle linear *inequalities*, also known as
linear programming. In contrast, if we insist on *integer* solutions,
the task of solving for linear equalities or inequalities is known as
integer
programming, and the
best known algorithms are exponential time in the worst case.

Whenever we discuss problems whose inputs correspond to numbers, the input length corresponds to how many bits are needed to describe the number (or, as is equivalent up to a constant factor, the number of digits in base 10, 16 or any other constant). The difference between the length of the input and the magnitude of the number itself can be of course quite profound. For example, most people would agree that there is a huge difference between having a billion (i.e. \(10^9\)) dollars and having nine dollars. Similarly there is a huge difference between an algorithm that takes \(n\) steps on an \(n\)-bit number and an algorithm that takes \(2^n\) steps.

One example, is the problem (discussed below) of finding the prime
factors of a given integer \(N\). The natural algorithm is to search for
such a factor by trying all numbers from \(1\) to \(N\), but that would take
\(N\) steps which is *exponential* in the input length, which is number of
bits needed to describe \(N\).The running time of this algorithm can be easily improved to
roughly \(\sqrt{N}\), but this is still exponential (i.e., \(2^{n/2}\))
in the number \(n\) of bits to describe \(N\). It is an important and long open
question whether there is such an algorithm that runs in time polynomial
in the input length (i.e., polynomial in \(\log N\)).

### Solving quadratic equations

Suppose that we want to solve not just *linear* but also equations
involving *quadratic* terms of the form \(a_{i,j,k}x_jx_k\). That is,
suppose that we are given a set of quadratic polynomials
\(p_1,\ldots,p_m\) and consider the equations \(\{ p_i(x) = 0 \}\). To avoid
issues with bit representations, we will always assume that the
equations contain the constraints \(\{ x_i^2 - x_i = 0 \}_{i\in [n]}\).
Since only \(0\) and \(1\) satisfy the equation \(a^2-a\), this assumption
means that we can restrict attention to solutions in \(\{0,1\}^n\).
Solving quadratic equations in several variable is a classical and
extremely well motivated problem. This is the generalization of the
classical case of single-variable quadratic equations that generations
of high school students grapple with. It also generalizes the quadratic
assignment
problem,
introduced in the 1950’s as a way to optimize assignment of economic
activities. Once again, we do not know a much better algorithm for this
problem than the one that enumerates over all the \(2^n\) possiblities.

## More advanced examples

We now list a few more examples of interesting problems that are a little more advanced but are of significant interest in areas such as physics, economics, number theory, and cryptography.

### The permanent (mod 2) problem

Given an \(n\times n\) matrix \(A\), the *permanent* of \(A\) is the sum over
all permutations \(\pi\) (i.e., \(\pi\) is a member of the set \(S_n\) of
one-to-one and onto functions from \([n]\) to \([n]\)) of the product
\(\prod_{i=0}^{n-1}A_{i,\pi(i)}\). The permanent of a matrix is a natural
quantity, and has been studied in several contexts including
combinatorics and graph theory. It also arises in physics where it can
be used to describe the quantum state of multiple boson particles (see
here and
here).

If the entries of \(A\) are integers, then we can also define a *Boolean*
function \(perm_2(A)\) which will output the result of the permanent
modulo \(2\). A priori computing this would seem to require enumerating
over all \(n!\) possiblities. However, it turns out we can compute
\(perm_2(A)\) in polynomial time! The key is that modulo \(2\), \(-x\) and
\(+x\) are the same quantity and hence the permanent moulo \(2\) is the same
as taking the following quantity modulo \(2\):

\[ \sum_{\pi \in S_n} sign(\pi)\prod_{i=0}^{n-1}A_{i,\pi(i)} \label{eq:det} \]

where the *sign* of a permutation \(\pi\) is a number in \(\{+1,-1\}\) which
can be defined in several ways, one of which is that \(sign(\pi)\) equals
\(+1\) if the number of swaps that “Bubble” sort performs starting an
array sorted according to \(\pi\) is even, and it equals \(-1\) if this
number is odd.It turns out that this definition is independent of the sorting
algorithm, and for example if \(sign(\pi)=-1\) then one cannot sort an
array ordered according to \(\pi\) using an even number of swaps.

From a first look, \eqref{eq:det} does not seem like it makes much progress. After all, all we did is replace one formula involving a sum over \(n!\) terms with an even more complicated formula involving a sum over \(n!\) tersm. But fortunately \eqref{eq:det} also has an alternative description: it is simply the determinant of the matrix \(A\), which can be computed using a process similar to Gaussian elimination.

### The permanent (mod 3) problem

Emboldened by our good fortune above, we might hope to be able to compute the permanent modulo any prime \(p\) and perhaps in full generality. Alas, we have no such luck. In a similar “two to three” type of a phenomenon, we do not know of a much better than brute force algorithm to even compute the permanent modulo \(3\).

### Finding a zero-sum equilibrium

A *zero sum game* is a game between two players where the payoff for one
is the same as the penalty for the other. That is, whatever the first
player gains, the second player loses. As much as we want to avoid them,
zero sum games do arise in life, and the one good thing about them is
that at least we can compute the optimal strategy.

A zero sum game can be specified by an \(n\times n\) matrix \(A\), where if
player 1 chooses action \(i\) and player 2 chooses action \(j\) then player
one gets \(A_{i,j}\) and player 2 loses the same amount. The famous Min
Max Theorem by John von
Neumann states that if we allow probabilistic or “mixed” strategies
(where a player does not choose a single action but rather a
*distribution* over actions) then it does not matter who plays first and
the end result will be the same. Mathematically the min max theorem is
that if we let \(\Delta_n\) be the set of probability distributions over
\([n]\) (i.e., non-negative columns vectors in \(\R^n\) whose entries sum to
\(1\)) then

\[ \max_{p \in \Delta_n} \min_{q\in \Delta_n} p^\top A q = \min_{q \in \Delta_n} \max_{p\in \Delta_n} p^\top A q \label{eq:minmax} \]

The min-max theorem turns out to be a corollary of linear programming duality, and indeed the value of \eqref{eq:minmax} can be computed efficiently by a linear program.

### Finding a Nash equilibrium

Fortunately, not all real-world games are zero sum, and we do have more
general games, where the payoff of one player does not necessarily qual
the loss of the other. John
Nash won the Nobel
prize for showing that there is a notion of *equilibrium* for such games
as well. In many economic texts it is taken as an article of faith that
when actual agents are involved in such a game then they reach a Nash
equilibrium. However, unlike zero sum games, we do not know of an
efficient algorithm for finding a Nash equilibrium given the description
of a general (non zero sum) game. In particular this means that, despite
economists’ intuitions, there are games for which natural stategies will
take exponential number of steps to converge to an equilibrium.

### Primality testing

Another classical computational problem, that has been of interest since
the ancient greeks, is to determine whether a given number \(N\) is prime
or composite. Clearly we can do so by trying to divide it with all the
numbers in \(2,\ldots,N-1\), but this would take at least \(N\) steps which
is *exponential* in its bit complexity \(n = \log N\). We can reduce these
\(N\) steps to \(\sqrt{N}\) by observing that if \(N\) is a composite of the
form \(N=PQ\) then either \(P\) or \(Q\) is smaller than \(\sqrt{N}\). But this
is still quite terrible. If \(N\) is a \(1024\) bit integer, \(\sqrt{N}\) is
about \(2^{512}\), and so running this algorithm on such an input would
take much more than the lifetime of the universe.

Luckily, it turns out we can do radically better. In the 1970’s, Rabin
and Miller gave *probabilistic* algorithms to determine whether a given
number \(N\) is prime or composite in time \(poly(n)\) for \(n=\log N\). We
will discuss the probabilistic model of computation later in this
course. In 2002, Agrawal, Kayal, and Saxena found a deterministic
\(poly(n)\) time algorithm for this problem. This is surely a development
that mathematicians from Archimedes till Gauss would have found
exciting.

### Integer factoring

Given that we can efficiently determine whether a number \(N\) is prime or
composite, we could expect that in the latter case we could also
efficiently *find* the factorization of \(N\). Alas, no such algorithm is
known. In a surprising and exciting turn of events, the *non existence*
of such an algorithm has been used as a basis for encryptions, and
indeed it underlies much of the security of the world wide web. We will
return to the factoring problem later in this course. We remark that we
do know much better than brute force algorithms for this problem. While
the brute force algorithms would require \(2^{\Omega(n)}\) time to factor
an \(n\)-bit integer, there are known algorithms running in time roughly
\(2^{O(\sqrt{n})}\) and also algorithms that are widely believed (though
not fully rigorously analyzed) to run in time roughly
\(2^{O(n^{1/3})}\).The “roughly” adjective above refers to neglecting factors that
are polylogarithmic in \(n\).

## Our current knowledge

The difference between an exponential and polynomial time algorithm might seem merely “quantiative” but it is in fact extremely significant. As we’ve already seen, the brute force exponential time algorithm runs out of steam very very fast, and as Edmonds says, in practice there might not be much difference between a problem where the best algorithm is exponential and a problem that is not solvable at all. Thus the efficient algorithms we mention above are widely used and power many computer science applications. Moreover, a polynomial-time algorithm often arises out of significant insight to the problem at hand, whether it is the “max-flow min-cut” result, the solvability of the determinant, or the group theoretic structure that enables primality testing. Such insight can be useful regardless of its computational implications.

At the moment we do not know whether the “hard” problems are truly hard,
or whether it is merely because we haven’t yet found the right
algorithms for them. However, we will now see that there are problems
that do *inherently require* exponential time. We just don’t know if any
of the examples above fall into that category.

## Lecture summary

- There are many natural problems that have polynomial-time algorithms, and other natural problems that we’d love to solve, but for which the best known algorithms are exponential.
- Often a polynomial time algorithm relies on discovering some hidden structure in the problem, or finding a surprising equivalent formulation for it.
- There are many interesting problems where there is an
*exponential gap*between the best known algorithm and the best algorithm that we can rule out. Closing this gap is one of the main open questions of theoretical computer science.

## Exercises

Give a \(poly(n)2^n\) time algorithm for the longest path problem in \(n\)
vertex graphs.**Hint:** Use dynamic programming to compute for every
\(s,t \in [n]\) and \(S \subseteq [n]\) the value \(P(s,t,S)\) which
equals to \(1\) if there is a simple path from \(s\) to \(t\) that uses
exactly the vertices in \(S\). Do this iteratively for \(S\)’s of
growing sizes.

For every 2CNF \(\varphi\), define the graph \(G_\varphi\) on \(2n\) vertices corresponding to the literals \(x_1,\ldots,x_n,\overline{x}_1,\ldots,\overline{x}_n\), such that there is an edge \(\overrightarrow{\ell_i\; \ell_j}\) iff the constraint \(\overline{\ell}_i \vee \ell_j\) is in \(\varphi\). Prove that \(\varphi\) is unsatisfiable if and only if there is some \(i\) such that there is a path from \(x_i\) to \(\overline{x}_i\) and from \(\overline{x}_i\) to \(x_i\) in \(G_\varphi\). Show how to use this to solve 2SAT in polynomial time.

A *regular* expression over the binary alphabet is a string consisting
of the symbols \(\{ 0 , 1 , \emptyset , ( , ) , * , | \}\). An expression
\(exp\) corresponds to a function mapping \(\{0,1\}^* \rightarrow \{0,1\}\),
where the `0`

and `1`

expressions correspond to the functions that map
only output \(1\) on the strings `0`

and `1`

respectively, and if
\(exp,exp'\) are expressions corresponnding to the functions, \(f'\) then
\((exp)|(exp')\) corresponds to the function \(f \vee f'\), \((exp)(exp')\)
corresponds to the function \(g\) such that for \(x\in \{0,1\}^n\), \(g(x)=1\)
if there is some \(i \in [n]\) such that \(f(x_0,\ldots,x_i)=1\) and
\(f(x_{i+1},\ldots,x_{n-1})=1\), and \((exp)*\) corresponds to the function
\(g\) such that \(g(x)=1\) if either \(x=\emptyset\) or there are strings
\(x_1,\ldots,x_k\) such that \(x\) is the concatenation of \(x_1,\ldots,x_k\)
and \(f(x_i)=1\) for every \(i\in [k]\).

Prove that for every regular expression \(exp\), the function
corresponding to \(exp\) is computable in polynomial time. Can you show
that it is computable in \(O(n)\) time?

## Bibliographical notes

Eugene Lawler’s quote on the “mystical power of twoness” was taken from the wonderful book “The Nature of Computation” by Moore and Mertens. See also this memorial essay on Lawler by Lenstra.

## Further explorations

Some topics related to this lecture that might be accessible to advanced students include: (to be completed)