Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.
One of the topics that seems to come up regularly on mailing lists and online discussions is the merits (or lack thereof) of doing a Computer Science Degree. An argument that seems to come up time and again for the negative party is that they have been coding for some number of years and they have never used recursion.
So the question is:
What is recursion?
When would I use recursion?
Why don't people use recursion?
There are a number of good explanations of recursion in this thread, this answer is about why you shouldn't use it in most languages.* In the majority of major imperative language implementations (i.e. every major implementation of C, C++, Basic, Python, Ruby,Java, and C#) iteration is vastly preferable to recursion.
To see why, walk through the steps that the above languages use to call a function:
space is carved out on the stack for the function's arguments and local variables
the function's arguments are copied into this new space
control jumps to the function
the function's code runs
the function's result is copied into a return value
the stack is rewound to its previous position
control jumps back to where the function was called
Doing all of these steps takes time, usually a little bit more than it takes to iterate through a loop. However, the real problem is in step #1. When many programs start, they allocate a single chunk of memory for their stack, and when they run out of that memory (often, but not always due to recursion), the program crashes due to a stack overflow.
So in these languages recursion is slower and it makes you vulnerable to crashing. There are still some arguments for using it though. In general, code written recursively is shorter and a bit more elegant, once you know how to read it.
There is a technique that language implementers can use called tail call optimization which can eliminate some classes of stack overflow. Put succinctly: if a function's return expression is simply the result of a function call, then you don't need to add a new level onto the stack, you can reuse the current one for the function being called. Regrettably, few imperative language-implementations have tail-call optimization built in.
* I love recursion. My favorite static language doesn't use loops at all, recursion is the only way to do something repeatedly. I just don't think that recursion is generally a good idea in languages that aren't tuned for it.
** By the way Mario, the typical name for your ArrangeString function is "join", and I'd be surprised if your language of choice doesn't already have an implementation of it.
Simple english example of recursion.
A child couldn't sleep, so her mother told her a story about a little frog,
who couldn't sleep, so the frog's mother told her a story about a little bear,
who couldn't sleep, so the bear's mother told her a story about a little weasel...
who fell asleep.
...and the little bear fell asleep;
...and the little frog fell asleep;
...and the child fell asleep.
In the most basic computer science sense, recursion is a function that calls itself. Say you have a linked list structure:
struct Node {
Node* next;
};
And you want to find out how long a linked list is you can do this with recursion:
int length(const Node* list) {
if (!list->next) {
return 1;
} else {
return 1 + length(list->next);
}
}
(This could of course be done with a for loop as well, but is useful as an illustration of the concept)
Whenever a function calls itself, creating a loop, then that's recursion. As with anything there are good uses and bad uses for recursion.
The most simple example is tail recursion where the very last line of the function is a call to itself:
int FloorByTen(int num)
{
if (num % 10 == 0)
return num;
else
return FloorByTen(num-1);
}
However, this is a lame, almost pointless example because it can easily be replaced by more efficient iteration. After all, recursion suffers from function call overhead, which in the example above could be substantial compared to the operation inside the function itself.
So the whole reason to do recursion rather than iteration should be to take advantage of the call stack to do some clever stuff. For example, if you call a function multiple times with different parameters inside the same loop then that's a way to accomplish branching. A classic example is the Sierpinski triangle.
You can draw one of those very simply with recursion, where the call stack branches in 3 directions:
private void BuildVertices(double x, double y, double len)
{
if (len > 0.002)
{
mesh.Positions.Add(new Point3D(x, y + len, -len));
mesh.Positions.Add(new Point3D(x - len, y - len, -len));
mesh.Positions.Add(new Point3D(x + len, y - len, -len));
len *= 0.5;
BuildVertices(x, y + len, len);
BuildVertices(x - len, y - len, len);
BuildVertices(x + len, y - len, len);
}
}
If you attempt to do the same thing with iteration I think you'll find it takes a lot more code to accomplish.
Other common use cases might include traversing hierarchies, e.g. website crawlers, directory comparisons, etc.
Conclusion
In practical terms, recursion makes the most sense whenever you need iterative branching.
Recursion is a method of solving problems based on the divide and conquer mentality.
The basic idea is that you take the original problem and divide it into smaller (more easily solved) instances of itself, solve those smaller instances (usually by using the same algorithm again) and then reassemble them into the final solution.
The canonical example is a routine to generate the Factorial of n. The Factorial of n is calculated by multiplying all of the numbers between 1 and n. An iterative solution in C# looks like this:
public int Fact(int n)
{
int fact = 1;
for( int i = 2; i <= n; i++)
{
fact = fact * i;
}
return fact;
}
There's nothing surprising about the iterative solution and it should make sense to anyone familiar with C#.
The recursive solution is found by recognising that the nth Factorial is n * Fact(n-1). Or to put it another way, if you know what a particular Factorial number is you can calculate the next one. Here is the recursive solution in C#:
public int FactRec(int n)
{
if( n < 2 )
{
return 1;
}
return n * FactRec( n - 1 );
}
The first part of this function is known as a Base Case (or sometimes Guard Clause) and is what prevents the algorithm from running forever. It just returns the value 1 whenever the function is called with a value of 1 or less. The second part is more interesting and is known as the Recursive Step. Here we call the same method with a slightly modified parameter (we decrement it by 1) and then multiply the result with our copy of n.
When first encountered this can be kind of confusing so it's instructive to examine how it works when run. Imagine that we call FactRec(5). We enter the routine, are not picked up by the base case and so we end up like this:
// In FactRec(5)
return 5 * FactRec( 5 - 1 );
// which is
return 5 * FactRec(4);
If we re-enter the method with the parameter 4 we are again not stopped by the guard clause and so we end up at:
// In FactRec(4)
return 4 * FactRec(3);
If we substitute this return value into the return value above we get
// In FactRec(5)
return 5 * (4 * FactRec(3));
This should give you a clue as to how the final solution is arrived at so we'll fast track and show each step on the way down:
return 5 * (4 * FactRec(3));
return 5 * (4 * (3 * FactRec(2)));
return 5 * (4 * (3 * (2 * FactRec(1))));
return 5 * (4 * (3 * (2 * (1))));
That final substitution happens when the base case is triggered. At this point we have a simple algrebraic formula to solve which equates directly to the definition of Factorials in the first place.
It's instructive to note that every call into the method results in either a base case being triggered or a call to the same method where the parameters are closer to a base case (often called a recursive call). If this is not the case then the method will run forever.
Recursion is solving a problem with a function that calls itself. A good example of this is a factorial function. Factorial is a math problem where factorial of 5, for example, is 5 * 4 * 3 * 2 * 1. This function solves this in C# for positive integers (not tested - there may be a bug).
public int Factorial(int n)
{
if (n <= 1)
return 1;
return n * Factorial(n - 1);
}
Recursion refers to a method which solves a problem by solving a smaller version of the problem and then using that result plus some other computation to formulate the answer to the original problem. Often times, in the process of solving the smaller version, the method will solve a yet smaller version of the problem, and so on, until it reaches a "base case" which is trivial to solve.
For instance, to calculate a factorial for the number X, one can represent it as X times the factorial of X-1. Thus, the method "recurses" to find the factorial of X-1, and then multiplies whatever it got by X to give a final answer. Of course, to find the factorial of X-1, it'll first calculate the factorial of X-2, and so on. The base case would be when X is 0 or 1, in which case it knows to return 1 since 0! = 1! = 1.
Consider an old, well known problem:
In mathematics, the greatest common divisor (gcd) … of two or more non-zero integers, is the largest positive integer that divides the numbers without a remainder.
The definition of gcd is surprisingly simple:
where mod is the modulo operator (that is, the remainder after integer division).
In English, this definition says the greatest common divisor of any number and zero is that number, and the greatest common divisor of two numbers m and n is the greatest common divisor of n and the remainder after dividing m by n.
If you'd like to know why this works, see the Wikipedia article on the Euclidean algorithm.
Let's compute gcd(10, 8) as an example. Each step is equal to the one just before it:
gcd(10, 8)
gcd(10, 10 mod 8)
gcd(8, 2)
gcd(8, 8 mod 2)
gcd(2, 0)
2
In the first step, 8 does not equal zero, so the second part of the definition applies. 10 mod 8 = 2 because 8 goes into 10 once with a remainder of 2. At step 3, the second part applies again, but this time 8 mod 2 = 0 because 2 divides 8 with no remainder. At step 5, the second argument is 0, so the answer is 2.
Did you notice that gcd appears on both the left and right sides of the equals sign? A mathematician would say this definition is recursive because the expression you're defining recurs inside its definition.
Recursive definitions tend to be elegant. For example, a recursive definition for the sum of a list is
sum l =
if empty(l)
return 0
else
return head(l) + sum(tail(l))
where head is the first element in a list and tail is the rest of the list. Note that sum recurs inside its definition at the end.
Maybe you'd prefer the maximum value in a list instead:
max l =
if empty(l)
error
elsif length(l) = 1
return head(l)
else
tailmax = max(tail(l))
if head(l) > tailmax
return head(l)
else
return tailmax
You might define multiplication of non-negative integers recursively to turn it into a series of additions:
a * b =
if b = 0
return 0
else
return a + (a * (b - 1))
If that bit about transforming multiplication into a series of additions doesn't make sense, try expanding a few simple examples to see how it works.
Merge sort has a lovely recursive definition:
sort(l) =
if empty(l) or length(l) = 1
return l
else
(left,right) = split l
return merge(sort(left), sort(right))
Recursive definitions are all around if you know what to look for. Notice how all of these definitions have very simple base cases, e.g., gcd(m, 0) = m. The recursive cases whittle away at the problem to get down to the easy answers.
With this understanding, you can now appreciate the other algorithms in Wikipedia's article on recursion!
A function that calls itself
When a function can be (easily) decomposed into a simple operation plus the same function on some smaller portion of the problem. I should say, rather, that this makes it a good candidate for recursion.
They do!
The canonical example is the factorial which looks like:
int fact(int a)
{
if(a==1)
return 1;
return a*fact(a-1);
}
In general, recursion isn't necessarily fast (function call overhead tends to be high because recursive functions tend to be small, see above) and can suffer from some problems (stack overflow anyone?). Some say they tend to be hard to get 'right' in non-trivial cases but I don't really buy into that. In some situations, recursion makes the most sense and is the most elegant and clear way to write a particular function. It should be noted that some languages favor recursive solutions and optimize them much more (LISP comes to mind).
A recursive function is one which calls itself. The most common reason I've found to use it is traversing a tree structure. For example, if I have a TreeView with checkboxes (think installation of a new program, "choose features to install" page), I might want a "check all" button which would be something like this (pseudocode):
function cmdCheckAllClick {
checkRecursively(TreeView1.RootNode);
}
function checkRecursively(Node n) {
n.Checked = True;
foreach ( n.Children as child ) {
checkRecursively(child);
}
}
So you can see that the checkRecursively first checks the node which it is passed, then calls itself for each of that node's children.
You do need to be a bit careful with recursion. If you get into an infinite recursive loop, you will get a Stack Overflow exception :)
I can't think of a reason why people shouldn't use it, when appropriate. It is useful in some circumstances, and not in others.
I think that because it's an interesting technique, some coders perhaps end up using it more often than they should, without real justification. This has given recursion a bad name in some circles.
Recursion is an expression directly or indirectly referencing itself.
Consider recursive acronyms as a simple example:
GNU stands for GNU's Not Unix
PHP stands for PHP: Hypertext Preprocessor
YAML stands for YAML Ain't Markup Language
WINE stands for Wine Is Not an Emulator
VISA stands for Visa International Service Association
More examples on Wikipedia
Recursion works best with what I like to call "fractal problems", where you're dealing with a big thing that's made of smaller versions of that big thing, each of which is an even smaller version of the big thing, and so on. If you ever have to traverse or search through something like a tree or nested identical structures, you've got a problem that might be a good candidate for recursion.
People avoid recursion for a number of reasons:
Most people (myself included) cut their programming teeth on procedural or object-oriented programming as opposed to functional programming. To such people, the iterative approach (typically using loops) feels more natural.
Those of us who cut our programming teeth on procedural or object-oriented programming have often been told to avoid recursion because it's error prone.
We're often told that recursion is slow. Calling and returning from a routine repeatedly involves a lot of stack pushing and popping, which is slower than looping. I think some languages handle this better than others, and those languages are most likely not those where the dominant paradigm is procedural or object-oriented.
For at least a couple of programming languages I've used, I remember hearing recommendations not to use recursion if it gets beyond a certain depth because its stack isn't that deep.
A recursive statement is one in which you define the process of what to do next as a combination of the inputs and what you have already done.
For example, take factorial:
factorial(6) = 6*5*4*3*2*1
But it's easy to see factorial(6) also is:
6 * factorial(5) = 6*(5*4*3*2*1).
So generally:
factorial(n) = n*factorial(n-1)
Of course, the tricky thing about recursion is that if you want to define things in terms of what you have already done, there needs to be some place to start.
In this example, we just make a special case by defining factorial(1) = 1.
Now we see it from the bottom up:
factorial(6) = 6*factorial(5)
= 6*5*factorial(4)
= 6*5*4*factorial(3) = 6*5*4*3*factorial(2) = 6*5*4*3*2*factorial(1) = 6*5*4*3*2*1
Since we defined factorial(1) = 1, we reach the "bottom".
Generally speaking, recursive procedures have two parts:
1) The recursive part, which defines some procedure in terms of new inputs combined with what you've "already done" via the same procedure. (i.e. factorial(n) = n*factorial(n-1))
2) A base part, which makes sure that the process doesn't repeat forever by giving it some place to start (i.e. factorial(1) = 1)
It can be a bit confusing to get your head around at first, but just look at a bunch of examples and it should all come together. If you want a much deeper understanding of the concept, study mathematical induction. Also, be aware that some languages optimize for recursive calls while others do not. It's pretty easy to make insanely slow recursive functions if you're not careful, but there are also techniques to make them performant in most cases.
Hope this helps...
I like this definition:
In recursion, a routine solves a small part of a problem itself, divides the problem into smaller pieces, and then calls itself to solve each of the smaller pieces.
I also like Steve McConnells discussion of recursion in Code Complete where he criticises the examples used in Computer Science books on Recursion.
Don't use recursion for factorials or Fibonacci numbers
One problem with
computer-science textbooks is that
they present silly examples of
recursion. The typical examples are
computing a factorial or computing a
Fibonacci sequence. Recursion is a
powerful tool, and it's really dumb to
use it in either of those cases. If a
programmer who worked for me used
recursion to compute a factorial, I'd
hire someone else.
I thought this was a very interesting point to raise and may be a reason why recursion is often misunderstood.
EDIT:
This was not a dig at Dav's answer - I had not seen that reply when I posted this
1.)
A method is recursive if it can call itself; either directly:
void f() {
... f() ...
}
or indirectly:
void f() {
... g() ...
}
void g() {
... f() ...
}
2.) When to use recursion
Q: Does using recursion usually make your code faster?
A: No.
Q: Does using recursion usually use less memory?
A: No.
Q: Then why use recursion?
A: It sometimes makes your code much simpler!
3.) People use recursion only when it is very complex to write iterative code. For example, tree traversal techniques like preorder, postorder can be made both iterative and recursive. But usually we use recursive because of its simplicity.
Here's a simple example: how many elements in a set. (there are better ways to count things, but this is a nice simple recursive example.)
First, we need two rules:
if the set is empty, the count of items in the set is zero (duh!).
if the set is not empty, the count is one plus the number of items in the set after one item is removed.
Suppose you have a set like this: [x x x]. let's count how many items there are.
the set is [x x x] which is not empty, so we apply rule 2. the number of items is one plus the number of items in [x x] (i.e. we removed an item).
the set is [x x], so we apply rule 2 again: one + number of items in [x].
the set is [x], which still matches rule 2: one + number of items in [].
Now the set is [], which matches rule 1: the count is zero!
Now that we know the answer in step 4 (0), we can solve step 3 (1 + 0)
Likewise, now that we know the answer in step 3 (1), we can solve step 2 (1 + 1)
And finally now that we know the answer in step 2 (2), we can solve step 1 (1 + 2) and get the count of items in [x x x], which is 3. Hooray!
We can represent this as:
count of [x x x] = 1 + count of [x x]
= 1 + (1 + count of [x])
= 1 + (1 + (1 + count of []))
= 1 + (1 + (1 + 0)))
= 1 + (1 + (1))
= 1 + (2)
= 3
When applying a recursive solution, you usually have at least 2 rules:
the basis, the simple case which states what happens when you have "used up" all of your data. This is usually some variation of "if you are out of data to process, your answer is X"
the recursive rule, which states what happens if you still have data. This is usually some kind of rule that says "do something to make your data set smaller, and reapply your rules to the smaller data set."
If we translate the above to pseudocode, we get:
numberOfItems(set)
if set is empty
return 0
else
remove 1 item from set
return 1 + numberOfItems(set)
There's a lot more useful examples (traversing a tree, for example) which I'm sure other people will cover.
Well, that's a pretty decent definition you have. And wikipedia has a good definition too. So I'll add another (probably worse) definition for you.
When people refer to "recursion", they're usually talking about a function they've written which calls itself repeatedly until it is done with its work. Recursion can be helpful when traversing hierarchies in data structures.
An example: A recursive definition of a staircase is:
A staircase consists of:
- a single step and a staircase (recursion)
- or only a single step (termination)
To recurse on a solved problem: do nothing, you're done.
To recurse on an open problem: do the next step, then recurse on the rest.
In plain English:
Assume you can do 3 things:
Take one apple
Write down tally marks
Count tally marks
You have a lot of apples in front of you on a table and you want to know how many apples there are.
start
Is the table empty?
yes: Count the tally marks and cheer like it's your birthday!
no: Take 1 apple and put it aside
Write down a tally mark
goto start
The process of repeating the same thing till you are done is called recursion.
I hope this is the "plain english" answer you are looking for!
A recursive function is a function that contains a call to itself. A recursive struct is a struct that contains an instance of itself. You can combine the two as a recursive class. The key part of a recursive item is that it contains an instance/call of itself.
Consider two mirrors facing each other. We've seen the neat infinity effect they make. Each reflection is an instance of a mirror, which is contained within another instance of a mirror, etc. The mirror containing a reflection of itself is recursion.
A binary search tree is a good programming example of recursion. The structure is recursive with each Node containing 2 instances of a Node. Functions to work on a binary search tree are also recursive.
This is an old question, but I want to add an answer from logistical point of view (i.e not from algorithm correctness point of view or performance point of view).
I use Java for work, and Java doesn't support nested function. As such, if I want to do recursion, I might have to define an external function (which exists only because my code bumps against Java's bureaucratic rule), or I might have to refactor the code altogether (which I really hate to do).
Thus, I often avoid recursion, and use stack operation instead, because recursion itself is essentially a stack operation.
You want to use it anytime you have a tree structure. It is very useful in reading XML.
Recursion as it applies to programming is basically calling a function from inside its own definition (inside itself), with different parameters so as to accomplish a task.
"If I have a hammer, make everything look like a nail."
Recursion is a problem-solving strategy for huge problems, where at every step just, "turn 2 small things into one bigger thing," each time with the same hammer.
Example
Suppose your desk is covered with a disorganized mess of 1024 papers. How do you make one neat, clean stack of papers from the mess, using recursion?
Divide: Spread all the sheets out, so you have just one sheet in each "stack".
Conquer:
Go around, putting each sheet on top of one other sheet. You now have stacks of 2.
Go around, putting each 2-stack on top of another 2-stack. You now have stacks of 4.
Go around, putting each 4-stack on top of another 4-stack. You now have stacks of 8.
... on and on ...
You now have one huge stack of 1024 sheets!
Notice that this is pretty intuitive, aside from counting everything (which isn't strictly necessary). You might not go all the way down to 1-sheet stacks, in reality, but you could and it would still work. The important part is the hammer: With your arms, you can always put one stack on top of the other to make a bigger stack, and it doesn't matter (within reason) how big either stack is.
Recursion is the process where a method call iself to be able to perform a certain task. It reduces redundency of code. Most recurssive functions or methods must have a condifiton to break the recussive call i.e. stop it from calling itself if a condition is met - this prevents the creating of an infinite loop. Not all functions are suited to be used recursively.
hey, sorry if my opinion agrees with someone, I'm just trying to explain recursion in plain english.
suppose you have three managers - Jack, John and Morgan.
Jack manages 2 programmers, John - 3, and Morgan - 5.
you are going to give every manager 300$ and want to know what would it cost.
The answer is obvious - but what if 2 of Morgan-s employees are also managers?
HERE comes the recursion.
you start from the top of the hierarchy. the summery cost is 0$.
you start with Jack,
Then check if he has any managers as employees. if you find any of them are, check if they have any managers as employees and so on. Add 300$ to the summery cost every time you find a manager.
when you are finished with Jack, go to John, his employees and then to Morgan.
You'll never know, how much cycles will you go before getting an answer, though you know how many managers you have and how many Budget can you spend.
Recursion is a tree, with branches and leaves, called parents and children respectively.
When you use a recursion algorithm, you more or less consciously are building a tree from the data.
In plain English, recursion means to repeat someting again and again.
In programming one example is of calling the function within itself .
Look on the following example of calculating factorial of a number:
public int fact(int n)
{
if (n==0) return 1;
else return n*fact(n-1)
}
Any algorithm exhibits structural recursion on a datatype if basically consists of a switch-statement with a case for each case of the datatype.
for example, when you are working on a type
tree = null
| leaf(value:integer)
| node(left: tree, right:tree)
a structural recursive algorithm would have the form
function computeSomething(x : tree) =
if x is null: base case
if x is leaf: do something with x.value
if x is node: do something with x.left,
do something with x.right,
combine the results
this is really the most obvious way to write any algorith that works on a data structure.
now, when you look at the integers (well, the natural numbers) as defined using the Peano axioms
integer = 0 | succ(integer)
you see that a structural recursive algorithm on integers looks like this
function computeSomething(x : integer) =
if x is 0 : base case
if x is succ(prev) : do something with prev
the too-well-known factorial function is about the most trivial example of
this form.
function call itself or use its own definition.
I've found many resources online, discussing this and related topics, but I haven't found anything that really helps me know where to start with implementing this solution.
To clarify, starting from city 0, I need to visit every other city once, and return back to city 0.
I have an array such as this one:
0 1129 1417 1240 1951
1129 0 1100 800 2237
1417 1100 0 1890 3046
1240 800 1890 0 1558
1951 2237 3046 1558 0
Along with finding the optimal route, I need to also find the optimal partial routes along the way. For example, I'd start with routes of length 2, and end up printing out something like this:
S = {0,1}
C({0,1},1) = 1129
S = {0,2}
C({0,2},2) = 1417
S = {0,3}
C({0,3},3) = 1240
S = {0,4}
C({0,4},4) = 1951
Then I'd go to routes of length 3, and print something like this:
S = {0,1,2}
C({0,1,2},1) = 2517
C({0,1,2},2) = 2229
and so on...
To make this a dynamic programming solution, I assume I should be saving the shortest distance between any n nodes, and the best way I've thought to do that is with a Hashmap, where the key would be an integer value of every node included in that path, in ascending order (A path going from nodes 0>1>3>4 or 0>1>4>3 could be stored as '134'), and each key would hold a pair that could store the path order as a List, and the total distance as an integer.
At this point I would think I'd want to calculate all paths of distance 2, then all of distance 3, and then take the smallest few and use the hashmap to find the shortest path back for each, and compare.
Does this seem like it could work? Or am I completely on the wrong track?
You're sort of on track. Dynamic programming isn't the way to calculate a TSP. What you're sort of close to is calculating a minimum spanning tree. This is a tree that connects all nodes using the shortest possible sum of edges. There are two algorithms that are frequently used: Primm's, and Kruskal's. They produce something similar to your optimal partial routes list. I'd recommend you look at Primm's algorithm: https://en.wikipedia.org/wiki/Prim%27s_algorithm
The easiest way of solving TSP is by finding the minimum spanning tree, and then doing a pre-order tree walk over the tree. This gives you an approximate travelling salesman solution, and is known as the Triangle Inequality Approximation. It's guaranteed to be no more than twice as long as an optimal TSP, but it can be calculated much faster. This web page explains it fairly well http://www.personal.kent.edu/~rmuhamma/Algorithms/MyAlgorithms/AproxAlgor/TSP/tsp.htm
If you want a more optimal solution, you'll need to look at Christofide's method, which is more complicated.
You are on the right track.
I think you're getting at the fact that the DP recursion itself only tells you the optimal cost for each S and j, not the node that attained that cost. The classical way to handle this is through "backtracking": once you found out that, for example, C({0,1,2,3,4},-) = 10000, you figure out which node attained that cost; let's say it's 3. Then you figure out which node attained C({0,1,2,3,4},3); let's say it's 1. Then you figure out which node attained C({0,1,2,4},1); let's say it's 2, ... and so on. This is the classical way and it avoids having to store a lot of intermediate data. For DPs with a small state space, it's easier just to store all those optimizers along the way. In your case you have an exponentially large state space so it might be expensive to store all of those optimizers, but you already have to store an equally large data structure (C), so most likely you can store the optimizers as well. Your intermediate solution of storing a few of the best could work -- then if the backtracking routine turns out to need one you didn't store you can just calculate it the classical way -- seems reasonable, but I'm not sure it's worth the extra coding vs. a pure backtracking approach.
One clarification: You don't actually want to "calculate all paths of distance 2, then all of distance 3, ...", but rather you want to enumerate all sets of size 2, then all sets of size 3, etc. Still exponential, but not as bad as enumerating all paths.
I have a problem in front of me that I can't grasp, even though I grasp the concept of backtracking well. I have been unable to find any information on this specific problem, and I am unaware if it goes by another name. The problem is to construct a backtracking algorithm that will print out the list of combinations of seating that N number of "presidents" can sit at a round table without sitting next to the same person twice. For example, you give it an input 8, it gives you the output,
1 2 3 4 5 6 7 8
1 3 5 2 6 8 4 7
1 4 2 7 5 8 3 6
Now I have constructed my own backtracking algorithms for simple games (checkers, n-queens, tic-tac-toe), however these that I have worked with have started with a blank canvas and didn't require this type of implementation. What I am asking for is a bit of guidance and insight, or links to more information to this problem as all that was given was about a page and a half, mostly talking about how the combination 1 2 3 4 is effectively the same as 1 4 3 2 in this particular problem. This needs to be done using a backtracking method and that's what is making me confused. If we were give free rein to find solutions I could easily do it. Any help is appreciated.
Thanks!
I'm thinking you'll have to add information about who sat next to who already to the backtracking algorithm state:
You start with, say 1. You seat 2 next to him, and record this "pairing" in the current partial solution. (Literally just an unordered pair of president IDs. An easy way of implementing an unordered pair of numbers is to have the constructor of that class store the smaller ID in the first element of the pair, and the greater ID in the second element.) So you have a partial solution 1 2 …, with the pairing 1-2.
When you reach to a complete solution, you add all its pairings to a "global" list of pairings. This will be used in future searches. I.e. if you get the solution 1 2 3 4 5 6 7 8, you save the pairings 1-2, 2-3, 3-4, 4-5, 5-6, 6-7, 7-8, 1-8.
Now, when doing a search, you have a partial solution x y z … – the last president seated so far is z, and the pairings in this partial solution are x-y and y-z. To seat the next president, you need to look at every president that's:
Not seated so far. (And thus not included in the partial solution's pairings.)
Not paired with z in the global pairings.
If no such president is available, you discard the partial solution and backtrack.
That said, I'm working off the top of my head here, so you'd be well advised to actually doublecheck if I missed any edge cases by using a straightforward brute force implementation and comparing the results.
A very rough pseudocode will look like this:
global flag // I know it is a bad practice but can't think of a better idea at this point
back (PresidentList, chairNumber, computedList)
if (PresidentList.size() == 0)
if (computedList doesn't violate any global distribution)
add computedList to globalList
add distirbution of computedList to global distirbution
set flag;
else return;
else if (flag is set and chairNumber > 2) return;
else if (flag is set and chairNumber = 2) un set flag; return;
else if (chairNumber == maxChairs-1)
// at this point there should be only one president in PresidentList
if (PresidentList.back() can sit with computedList.back() and computedList.front())
remove president from PreseidentList, add president to computedList
back(PresidenList, chairNumber+1,computedList)
else return;
else
for every president in PresidentList
if (president can sit with computedList.back())
remove president from PreseidentList, add president to computedList
back(PresidentList,chairNumber+1,computedList);
You always check if a president can sit near some other president only according to the global distribution variable.
In the main function just run a loop for PresidentList that will start back with that president being as the first one.
The idea of the flag is as follows: when you add a distribution and a list to the global list, back runs already a lot of branches which have the same prefix as the list you just added and hence are already bad. To close all those lists you force the back to stop branching and return to the point when it can set a new 2nd element in the list to compute i.e. get back to the loop if the last else branch of the function.
I've got a lot of true/false results saved as bits in long[] arrays. I do have a huge number of these (millions and millions of longs).
For example, say I have only five results, I'd have:
+----- condition 5 is true
|
|+---- condition 4 is false
||
||+--- condition 3 is true
|||
|||+-- condition 2 is true
||||
||||+- condition 1 is false
10110
I also do have a few trees representing statements like:
condition1 AND (condition2 OR (condition3 AND condition 4))
The trees are very simple but very long. They basically look like this (it's an oversimplification below, just to show what I've got):
class Node {
int operator();
List<Node> nodes;
int conditionNumber();
}
Basically either the Node is a leaf and then has a condition number (matching one of the bit in the long[] arrays) or the Node is not a leaf and hence refers several subnodes.
They're simple yet they allow to express complicated boolean expressions. It works great.
So far so good, everything is working great. However I do have a problem: I need to evaluate a LOT of expressions, determining if they're true or false. Basically I need to do some brute-force computation for a problem for which there's no know better solution than brute-forcing.
So I need to walk the tree and answer either true or false depending on the content of the tree and the content of the long[].
The method I need to optimize looks like this:
boolean solve( Node node, long[] trueorfalse ) {
...
}
where on the first call the node is the root node and then, obviously, subnodes (being recursive, that solve method calls itself).
Knowing that I'll only have a few trees (maybe up to a hundred or so) but millions and millions of long[] to check, what steps can I take to optimize this?
The obvious recursive solution passes parameters (the (sub)tree and the long[], I could get rid of the long[] by not passing it as a parameter) and is quite slow with all the recursive calls etc. I need to check which operator is used (AND or OR or NOT etc.) and there's quite a lot of if/else or switch statements involved.
I'm not looking for another algorithm (there aren't) so I'm not looking for going from O(x) to O(y) where y would be smaller than x.
What I'm looking for is "times x" speedup: if I can write code performing 5x faster, then I'll have a 5x speedup and that's it and I'd be very happy with it.
The only enhancement I see as of now --and I think it would be a huge "times x" speedup compared to what I have now-- would be to generate bytecode for every tree and have the logic for every tree hardcoded into a class. It should work well because I'll only ever have a hundred or so trees (but the trees aren't fixed: I cannot know in advance how the trees are going to look like, otherwise it would be trivial to simply hardcode manually every tree).
Any idea besides generating bytecode for every tree?
Now if I want to try the bytecode generation route, how should I go about it?
In order to maximize the opportunities for shortcut evaluation, you need to do your own branch prediction.
You might want to profile it, tallying
which AND branches evaluate into false
which OR branches result into true
You can then reorder the tree relative to the weights that you found in the profiling step. If you want/need to be particularly nifty, you can devise a mechanism that detects the weighting for a certain dataset during runtime, so you can reorder the branches on the fly.
Note that in the latter case, it might be advisable to not reorder the actual tree (with respect to storage efficiency and correctness of result while still executing), but rather devise a tree-node visitor (traversal algorithm) that is able to locally sort the branches according to the 'live' weights.
I hope all of this makes sense, because I realize the prose version is dense. However, like Fermat said, the code example is too big to fit into this margin :)
There is a simple and fast way to evaluate boolean operations like this in C. Assuming you want to evaluate z=(x op y) you can do this:
z = result[op+x+(y<<1)];
So op will be a multiple of 4 to select your operation AND, OR, XOR, etc. you create a lookup table for all possible answers. If this table is small enough, you can encode it into a single value and use right shift and a mask to select the output bit:
z = (MAGIC_NUMBER >> (op+x+(y<<1))) & 1;
That would be the fastest way to evaluate large numbers of these. Of course you'll have to split operations with multiple inputs into trees where each node has only 2 inputs. There is no easy way to short circuit this however. You can convert the tree into a list where each item contains the operation number and pointers to the 2 inputs and output. Once in list form, you can use a single loop to blow through that one line a million times very quickly.
For small trees, this is a win. For larger trees with short circuiting it's probably not a win because the average number of branches that need to be evaluated goes from 2 to 1.5 which is a huge win for large trees. YMMV.
EDIT:
On second thought, you can use something like a skip-list to implement short circuiting. Each operation (node) would include a compare value and a skip-count. if the result matched the compare value, you can bypass the next skip-count values. So the list would be created from a depth-first traversal of the tree, and the first child would include a skip count equal to the size of the other child. This takes a bit more complexity to each node evaluation but allows short circuiting. Careful implementation could do it without any condition checking (think 1 or 0 times the skip-count).
I think your byte-coding idea is the right direction.
What I would do, regardless of language, is write a precompiler.
It would walk each tree, and use print statements to translate it into source code, such as.
((word&1) && ((word&2) || ((word&4) && (word&8))))
That can be compiled on the fly whenever the trees change, and the resulting byte code / dll loaded, all of which takes under a second.
The thing is, at present you are interpreting the contents of the trees.
Turning them into compiled code should make them run 10-100 times faster.
ADDED in response to your comments on not having the JDK. Then, if you can't generate Java byte code, I would try to write my own byte-code interpreter than would run as fast as possible. It could look something like this:
while(iop < nop){
switch(code[iop++]){
case BIT1: // check the 1 bit and stack a boolean
stack[nstack++] = ((word & 1) != 0);
break;
case BIT2: // check the 2 bit and stack a boolean
stack[nstack++] = ((word & 2) != 0);
break;
case BIT4: // check the 4 bit and stack a boolean
stack[nstack++] = ((word & 4) != 0);
break;
// etc. etc.
case AND: // pop 2 booleans and push their AND
nstack--;
stack[nstack-1] = (stack[nstack-1] && stack[nstack]);
break;
case OR: // pop 2 booleans and push their OR
nstack--;
stack[nstack-1] = (stack[nstack-1] || stack[nstack]);
break;
}
}
The idea is to cause the compiler to turn the switch into a jump table, so it performs each operation with the smallest number of cycles.
To generate the opcodes, you can just do a postfix walk of the tree.
On top of that, you might be able to simplify it by some manipulation of De Morgan's laws, so you could check more than one bit at a time.