Local Computation Algorithms
summary
Defines a formal model where an algorithm answers queries about pieces of a huge output without computing the whole output, proving that many tasks admit logarithmic time or polylogarithmic query and space local algorithms with failure probability ≈1/n. Techniques include simulating constant‑round distributed algorithms, using shared random seed for consistency, and running O(d) rounds of Luby to break the graph into polylog‑sized components then brute‑force locally. Shows MIS, local decoding/decompression, and monotonicity correction work; dependence on degree remains the main open bottleneck.
AI-written summary · about