What is the best (regarding performance) way to compute the critical path of a directional acyclic graph when the nodes of the graph have weight?
For example, if I have the following structure:
Node A (weight 3)
/ \
Node B (weight 4) Node D (weight 7)
/ \
Node E (weight 2) Node F (weight 3)
The critical path should be A->B->F (total weight: 10)
I would solve this with dynamic programming. To find the maximum cost from S to T:
Topologically sort the nodes of the graph as S = x_0, x_1, ..., x_n = T. (Ignore any nodes that can reach S or be reached from T.)
The maximum cost from S to S is the weight of S.
Assuming you've computed the maximum cost from S to x_i for all i < k, the maximum cost from S to x_k is the cost of x_k plus the maximum cost to any node with an edge to x_k.
I have no clue about "critical paths", but I assume you mean this.
Finding the longest path in an acyclic graph with weights is only possible by traversing the whole tree and then comparing the lengths, as you never really know how the rest of the tree is weighted. You can find more about tree traversal at Wikipedia. I suggest, you go with pre-order traversal, as it's easy and straight forward to implement.
If you're going to query often, you may also wish to augment the edges between the nodes with information about the weight of their subtrees at insertion. This is relatively cheap, while repeated traversal can be extremely expensive.
But there's nothing to really save you from a full traversal if you don't do it. The order doesn't really matter, as long as you do a traversal and never go the same path twice.
There's a paper that purports to have an algorithm for this: "Critical path in an activity network with time constraints". Sadly, I couldn't find a link to a free copy. Short of that, I can only second the idea of modifying http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm or http://en.wikipedia.org/wiki/A*
UPDATE: I apologize for the crappy formatting—the server-side markdown engine is apparently broken.
My first answer so please excuse for any non-standard thing by the culture of stackoverflow.
I think the solution is simple. Just negate the weights and run the classic shortest path for DAG (modified for weights of vertices of course). It should run fairly fast. (Time complexity of O(V+E) maybe)
I think it should work as when you will negate the weights, the biggest one will become smallest, second biggest will be second smallest and so on as if a > b then -a < -b. Then running DAG should suffice as it will find the solution for the smallest path of the negated one and thus finding longest path for the original one
Try the A* method.
A* Search Algorithm
At the end, to deal with the leaves, just make all of them lead on to a final point, to set as the goal.
Related
I was asked the following question in an interview and I am unable to find an efficient solution.
Here is the problem:
We want to build a network and we are given c nodes/cities and D possible edges/connections made by roads. Edges are bidirectional and we know the cost of the edge. The costs of the edges can be represented as d[i,j] which denotes the cost of the edge i-j. Note not all c nodes can be directly connected to each other (D is the set of possible edges).
Now we are given a list of k potential edges/connections that have no cost. However, you can only choose one edge in the list of k edges to use (like getting free funding to build an airport between two cities).
So the question is... find the set of roads (and the one free airport) that minimizes total cost required to build the network connecting all cities in an efficient runtime.
So in short, solve a minimum spanning tree problem but where you can choose 1 edge in a list of k potential edges to be free of cost. I'm unsure how to solve... I've tried finding all the spanning trees in order of increasing cost and choosing the lowest cost, but I'm still challenged on how to consider the one free edge from the list of k potential free edges. I've also tried finding the MST of the D potential connections and then adjusting it according the the options in k to get a result.
Thank you for any help!
One idea would be to treat your favorite MST algorithm as a black box and to think about changing the edges in the graph before asking for the MST. For example, you could try something like this:
for each edge in the list of possible free edges:
make the graph G' formed by setting that edge cost to 0.
compute the MST of G'
return the cheapest MST out of all the ones generated this way
The runtime of this approach is O(kT(m, n)), where k is the number of edges to test and T(m, n) is the cost of computing an MST using your favorite black-box algorithm.
We can do better than this. There's a well-known problem of the following form:
Suppose you have an MST T for a graph G. You then reduce the cost of some edge {u, v}. Find an MST T' in the new graph G'.
There are many algorithms for solving this problem efficiently. Here's one:
Run a DFS in T starting at u until you find v.
If the heaviest edge on the path found this way costs more than {u, v}:
Delete that edge.
Add {u, v} to the spanning tree.
Return the resulting tree T'.
(Proving that this works is tedious but doable.) This would give an algorithm of cost O(T(m, n) + kn), since you would be building an initial MST (time T(m, n)), then doing k runs of DFS in a tree with n nodes.
However, this can potentially be improved even further if you're okay using some more advanced algorithms. The paper "On Cartesian Trees and Range Minimum Queries" by Demaine et al shows that in O(n) time, it is possible to preprocess a minimum spanning tree so that, in time O(1), queries of the form "what is the lowest-cost edge on the path in this tree between nodes u and v?" in time O(1). You could therefore build this structure instead of doing a DFS to find the bottleneck edge between u and v, reducing the overall runtime to O(T(m, n) + n + k). Given that T(m, n) is very low (the best known bound is O(m α(m)), where α(m) is the Ackermann inverse function and is less than five for all inputs in the feasible univers), this is asymptotically a very quick algorithm!
First generate a MST. Now, if you add a free edge, you will create exactly one cycle. You could then remove the heaviest edge in the cycle to get a cheaper tree.
To find the best tree you can make by adding one free edge, you need to find the heaviest edge in the MST that you could replace with a free one.
You can do that by testing one free edge at a time:
Pick a free edge
Find the lowest common ancestor in the tree (from an arbitrary root) of its adjacent vertices
Remember the heaviest edge on the path between the free edge vertices
When you're done, you know which free edge to use -- it's the one associated with the heaviest tree edge, and you know which edge it replaces.
In order to make steps (2) and (3) faster, you can remember the depth of each node and connect it to multiple ancestors like a skip list. You can then do those steps in O(log |V|) time, leading to a total complexity of O( (|E|+k) log |V| ), which is pretty good.
EDIT: Even Easier Way
After thinking about this a bit, it seems there's a super easy way to figure out which free edge to use and which MST edge to replace.
Disregarding the k possible free edges, you build the MST from the other edges using Kruskal's algorithm, but you modify the usual disjoint set data structure as follows:
Use union by size or rank, but not path compression. Every union operation will then establish exactly one link, and take O(log N) time, and all path lengths will be at most O(log N) long.
For each link, remember the index of the edge that caused it to be created.
For each possible free edge, then, you can walk up the links in the disjoint set structure to find out exactly at which point its endpoints were connected into the same connected component. You get the index of the last required edge, i.e., the one it would replace, and the free edge with the greatest replacement target index is the one you should use.
The title is self explanatory. Here's a solution that I found in the internet that can help do this. Here's the link
I don't understand why not visiting a vertex having weight below the given threshold will solve the problem.
Additionally, I have no idea how to solve this using/not using this.
Let's restrict this to simple cycles - those which contain no subcycles. For each node in the graph, begin a depth-first search for that node. record each branch of the recursion tree which results in a match. While searching, never cross over nodes already traversed in the branch.
Consider the complete directed graph on n vertices. There are n(n-1) arcs and n! simple cycles of length n. The algorithm above isn't much worse than this at all. Simply constructing a new copy of the answer would take nearly as much time as running the above algorithm to do it, in the worst case at least.
If you want to find cycles in a directed (or even undirected) graph there is an intuitive way to do it:
For each edge (u, v) in the graph
1. Temporarily ignore the edge (u, v) in step 2
2. Run an algorithm to find all paths from v to u (using a backtrackig algorithm)
3. Output the computed paths in step 2 along with the edge (u, v) as cycles in the graph
Note that you will get duplicate cycles this way since a cycle of length k will be found k times.
You can play with this idea to find cycles with specific properties, as well. For example, if you are aiming to find the shortest weighted cycle in the graph instead of finding all cycles. You can use a Dijkstra in step 2, and take the minimum over all the cycles you find. If you wanna finding the cycle with the least number of edges you can use a BFS in step 2.
If you are more struggling with finding all paths in a graph, this question might help you. Although it's for a slightly different problem.
Counting/finding paths with backtracking
It is said that DFS can't be used to find the shortest path in the unweighted graph. I have read multiple post and blogs but not get satisfied as a little modification in DFS can make it possible.
I think if we use Modified DFS in this way, then we can find the shortest distances from the source.
Initialise a array of distances from root with infinity and distance of root from itself as 0.
While traversing, we keep track of no. of edges. On moving forward increment no. of edges and while back track decrement no. of edges. And each time check if(dist(v) > dist(u) + 1 ) then dist(v) = dist(u) + 1.
In this way we can find the shortest distances from the root using DFS. And in this way, we can find it in O(V+E) instead of O(ElogV) by Dijkstra.
If I am wrong at some point. Please tell me.
Yes, if the DFS algorithm is modified in the way you mentioned, it can be used to find the shortest paths from a root in an unweighted graph. The problem is that in modifying the algorithm you have fundamentally changed what it is.
It may seem like I am exaggerating as the change looks minor superficially but it changes it more than you might think.
Consider a graph with n nodes numbered 1 to n. Let there be an edge between each k and k + 1. Also, let 1 be connected to every node.
Since DFS can pick adjacent neighbors in any order, let's also assume that this algorithm always picks them in increasing numerical order.
Now try running algorithm in your head or your computer with root 1.
First the algorithm will reach n in n-1 steps using edges between 1-2, 2-3 and so on. Then after backtracking, the algorithm moves on to the second neighbor of 1, namely 3. This time there will be n-2 steps.
The same process will repeat until the algorithm finally sees 1-n.
The algorithm will need O(n ^ 2) rather than O(n) steps to finish. Remember that V = n & E = 2 * n - 3. So it is not O(V + E).
Actually, the algorithm you have described will always finish in O(V^2) on unweighted graphs. I will leave the proof of this claim as an exercise for the reader.
O(V^2) is not that bad. Especially if a graph is dense. But since BFS already provides an answer in O(V + E), nobody uses DFS for shortest distance calculation.
In an unweighted graph, you can use a breadth-first search (not DFS) to find shortest paths in O(E) time.
In fact, if all edges have the same weight, then Dijkstra's algorithm and breadth-first search are pretty much equivalent -- reduceKey() is never called, and the priority queue can be replaced with a FIFO queue, since newly added vertices never have smaller weight than previously-added ones.
Your modification to DFS does not work, because once you visit a vertex, you will not examine its children again, even if its weight changes. You will get the wrong answer for this graph if you follow S->A before S->B
S---->A---->C---->D---->E---->T
\ /
------->B-----/
The way Depth First Search on graphs is defined, it only visits each node once. When it encounters a node that was visited before, it backtracks.
So assume you have a triangle with nodes A, B, C and you want to find the shortest path from A to B. One possibility of a DFS traversal is A -> C -> B and you are done. This however is not the shortest path.
I have a large graph with millions of nodes. I want to check if node 'A' is reachable from node 'B' with less than 4 hops. If possible, I want the shortest path. Which is the best way (or algorithm) to solve this issue?
Note that if the graph is unweighted (as it seems in your question) - a simple and efficient BFS will be enough to find the shortest path from the source to the target.
Also, since you have a single source and a single target - you can apply bi-directional BFS, which is more efficient then BFS.
Algorithm idea: do a BFS search simultaneously from the source and the target: [BFS until depth 1 in both, until depth 2 in both, ....].
The algorithm will end when you find a vertex v, which is in both BFS's front.
Algorithm behavior: The vertex v that terminates the algorithm's run will be exactly in the middle between the source and the target.
This algorithm will yield much better result in most cases then BFS from the source [explanation why it is better then BFS follows], and will surely provide an answer, if one exist.
why is it better then BFS from the source?
assume the distance between source to target is k, and the branch factor is B [every vertex has B edges].
BFS will open: 1 + B + B^2 + ... + B^k vertices.
bi-directional BFS will open: 2 + 2B^2 + 2B^3 + .. + 2B^(k/2) vertices.
for large B and k, the second is obviously much better the the first.
(*) The explanation of bi-directional search is taken from another answer I posted
The best algorithm for finding the shortest path between two nodes in a graph in which you have no extra information about how likely it is that one node is close to the target is Dijkstra's Algorithm. You can easily modify this algorithm to quit after 3 hops, to avoid wasting computation on results that you are not interested in.
If you do have some extra information about the likelihood that a given node is close to your target, you can use A* search, which uses a heuristic on a node's distance to its target to improve its runtime performance.
If you need path less than 3 hops, then all possible paths are A (A=B), A-B (nodes are adjacent), A-X-B (X is a node adjacent to both ends). So there's no need in any complex algorithm. First, test for A=B, second, test that A and B are adjacent, and third, try to find X that is adjacent to both A and B (eg. intesection of endpoint adjacency sets).
I have Graph with N nodes and edges with cost. (graph may be Complete but also can contain zero edges).
I want to find K trees in the graph (K < N) to ensure every node is visited and cost is the lowest possible.
Any recommendations what the best approach could be?
I tried to modify the problem to finding just single minimal spanning tree, but didn't succeeded.
Thank you for any hint!
EDIT
little detail, which can be significant. To cost is not related to crossing the edge. The cost is the price to BUILD such edge. Once edge is built, you can traverse it forward and backwards with no cost. The problem is not to "ride along all nodes", the problem is about "creating a net among all nodes". I am sorry for previous explanation
The story
Here is the story i have heard and trying to solve.
There is a city, without connection to electricity. Electrical company is able to connect just K houses with electricity. The other houses can be connected by dropping cables from already connected houses. But dropping this cable cost something. The goal is to choose which K houses will be connected directly to power plant and which houses will be connected with separate cables to ensure minimal cable cost and all houses coverage :)
As others have mentioned, this is NP hard. However, if you're willing to accept a good solution, you could use simulated annealing. For example, the traveling salesman problem is NP hard, yet near-optimal solutions can be found using simulated annealing, e.g. http://www.codeproject.com/Articles/26758/Simulated-Annealing-Solving-the-Travelling-Salesma
You are describing something like a cardinality constrained path cover. It's in the Traveling Salesman/ Vehicle routing family of problems and is NP-Hard. To create an algorithm you should ask
Are you only going to run it on small graphs.
Are you only going to run it on special cases of graphs which do have exact algorithms.
Can you live with a heuristic that solves the problem approximately.
Assume you can find a minimum spanning tree in O(V^2) using prim's algorithm.
For each vertex, find the minimum spanning tree with that vertex as the root.
This will be O(V^3) as you run the algorithm V times.
Sort these by total mass (sum of weights of their vertices) of the graph. This is O(V^2 lg V) which is consumed by the O(V^3) so essentially free in terms of order complexity.
Take the X least massive graphs - the roots of these are your "anchors" that are connected directly to the grid, as they are mostly likely to have the shortest paths. To determine which route it takes, you simply follow the path to root in each node in each tree and wire up whatever is the shortest. (This may be further optimized by sorting all paths to root and using only the shortest ones first. This will allow for optimizations on the next iterations. Finding path to root is O(V). Finding it for all V X times is O(V^2 * X). Because you would be doing this for every V, you're looking at O(V^3 * X). This is more than your biggest complexity, but I think the average case on these will be small, even if their worst case is large).
I cannot prove that this is the optimal solution. In fact, I am certain it is not. But when you consider an electrical grid of 100,000 homes, you can not consider (with any practical application) an NP hard solution. This gives you a very good solution in O(V^3 * X), which I imagine is going to give you a solution very close to optimal.
Looking at your story, I think that what you call a path can be a tree, which means that we don't have to worry about Hamiltonian circuits.
Looking at the proof of correctness of Prim's algorithm at http://en.wikipedia.org/wiki/Prim%27s_algorithm, consider taking a minimum spanning tree and removing the most expensive X-1 links. I think the proof there shows that the result has the same cost as the best possible answer to your problem: the only difference is that when you compare edges, you may find that the new edge join two separated components, but in this case you can maintain the number of separated components by removing an edge with cost at most that of the new edge.
So I think an answer for your problem is to take a minimum spanning tree and remove the X-1 most expensive links. This is certainly the case for X=1!
Here is attempt at solving this...
For X=1 I can calculate minimal spanning tree (MST) with Prim's algorithm from each node (this node is the only one connected to the grid) and select the one with the lowest overall cost
For X=2 I create extra node (Power plant node) beside my graph. I connect it with random node (eg. N0) by edge with cost of 0. I am now sure I have one power plant plug right (the random node will definitely be in one of the tree, so whole tree will be connected). Now the iterative part. I take other node (eg. N1) and again connected with PP with 0 cost edge. Now I calculate MST. Then repeat this process with replacing N1 with N2, N3 ...
So I will test every pair [N0, NX]. The lowest cost MST wins.
For X>2 is it really the same as for X=2, but I have to test connect to PP every (x-1)-tuple and calculate MST
with x^2 for MST I have complexity about (N over X-1) * x^2... Pretty complex, but I think it will give me THE OPTIMAL solution
what do you think?
edit by random node I mean random but FIXED node
attempt to visualize for x=2 (each description belongs to image above it)
Let this be our city, nodes A - F are houses, edges are candidates to future cables (each has some cost to build)
Just for image, this could be the solution
Let the green one be the power plant, this is how can look connection to one tree
But this different connection is really the same (connection to power plant(pp) cost the same, cables remains untouched). That is why we can set one of the nodes as fixed point of contact to the pp. We can be sure, that the node will be in one of the trees, and it does not matter where in the tree is.
So let this be our fixed situation with G as PP. Edge (B,G) with zero cost is added.
Now I am trying to connect second connection with PP (A,G, cost 0)
Now I calculate MST from the PP. Because red edges are the cheapest (the can actually have even negative cost), is it sure, that both of them will be in MST.
So when running MST I get something like this. Imagine detaching PP and two MINIMAL COST trees left. This is the best solution for A and B are the connections to PP. I store the cost and move on.
Now I do the same for B and C connections
I could get something like this, so compare cost to previous one and choose the better one.
This way I have to try all the connection pairs (B,A) (B,C) (B,D) (B,E) (B,F) and the cheapest one is the winner.
For X=3 I would just test other tuples with one fixed again. (A,B,C) (A,B,D) ... (A,C,D) ... (A,E,F)
I just came up with the easy solution as follows:
N - node count
C - direct connections to the grid
E - available edges
1, Sort all edges by cost
2, Repeat (N-C) times:
Take the cheapest edge available
Check if adding this edge will not caused circles in already added edge
If not, add this edge
3, That is all... You will end up with C disjoint sets of edges, connect every set to the grid
Sounds like the famous Traveling Salesman problem. The problem known to be NP-hard. Take a look at the Wikipedia as your starting point: http://en.wikipedia.org/wiki/Travelling_salesman_problem