Faster and Simpler Algorithms for Multicommodity Flow and Other Fractional Packing Problems
read more
Citations
The Multiplicative Weights Update Method: A Meta-Algorithm and Applications
Bandits with Knapsacks
Optimal content placement for a large-scale VoD system
Bandits with Knapsacks
Lifetime maximization for connected target coverage in wireless sensor networks
References
Fast approximation algorithms for fractional packing and covering problems
The maximum concurrent flow problem
Speeding-up linear programming using fast matrix multiplication
Improved approximation algorithms for the multi-commodity flow problem and local competitive routing in dynamic networks
Related Papers (5)
Fast approximation algorithms for fractional packing and covering problems
The Multiplicative Weights Update Method: A Meta-Algorithm and Applications
Frequently Asked Questions (7)
Q2. What is the dual of this linear program?
The dual of this linear program is an assignment of lengths to the edges, l : E ! R + , and a scalar | which the authors view as a length associated with a pseudo-edge of capacity B | such thatD(l; )def =Pel(e)c(e)+
Q3. What was the key idea of the algorithm?
The key idea of their procedure, which was adopted in a lot of subsequent work, was to compute an initial ow by disregarding edge capacities and then to reroute this, iteratively, along short paths so as to reduce the maximum congestion on any edge.
Q4. What is the minimum capacity edge on this path?
If the minimum capacity edge on this pathhas capacity c then the ow function at this step, fs i;j , corresponds to routing c units of ow alongthis path.
Q5. What is the simplest way to increase the LHS of the i th constraint?
By their de nition of p it follows that z 1 and hence increasing the LHS of the i th constraint by 1 causes an increase in y(i) by a multiplicative factor of at least 1 + .
Q6. What is the simplest way to calculate the ow problem?
The dual to the min-cost multicommodity ow problemis an assignment of lengths to edges, l : E ! R+, and a scalar such that D(l)= (l) is minimized.
Q7. What is the bound for ow?
In particular, for multicommodity ow, it implies a procedure which does not involve rerouting ow (the ow is only scaled at the end) and which for the case of maximum s-t ow reduces to the algorithm discussed at the beginning of this section.