← ClaudeAtlas

algorithm-design-paradigmslisted

Algorithm design from a problem statement: formalise, recognise the classic problem in disguise, pick a paradigm (greedy, divide and conquer, search, solver, streaming, online) and meet its proof obligation, then verify with a brute-force oracle and differential tests. Load when designing an algorithm or judging whether a proposed approach is correct.
rustcircle/c101 · ★ 0 · Web & Frontend · score 65
Install: claude install-skill rustcircle/c101
# Algorithm design paradigms From statement to verified algorithm. DP technique lives in `dynamic-programming-patterns`, graphs in `graph-algorithms`, proofs of bounds in `amortized-recurrence-analysis`, hardness in `np-hardness-reductions`, randomness and sketches in `randomized-algorithms-sketches`; the hub is `cs-foundations` section 1. ## 1. Formalisation template Write this down before choosing anything: | Field | Fill in | |---|---| | Input | types, n, m, value ranges (weights up to 10^9? negative? reals?) | | Output | value only, or the witness (set, order, path) too | | Objective | min / max / count (mod p?) / decide / enumerate all | | Constraints | hard vs soft; ties; online or all input known; queries after preprocessing | | Machine model | word RAM (w = 64), comparison only, external memory (B, M), work-span (W, D), streaming (one pass, sublinear space) | | Budget | time limit, memory limit, latency percentile if it serves requests | Size-to-budget table at ~10^8 simple operations per second per core (C++/Rust/Java; divide n by roughly 10-50 for pure Python loops, or vectorise): | n up to | Affordable | Typical algorithm | |---|---|---| | 10 | n! (3.6e6) | permutations, brute force | | 20 | 2^n · n (2e7) | bitmask DP, subset enumeration | | 40 | 2^(n/2) · n | meet in the middle | | 500 | n^3 (1.25e8) | Floyd-Warshall, interval DP | | 5 000 | n^2 (2.5e7) | pairwise DP, O(n^2) DP | | 10^6 | n log n (2e7) | sorting, heaps, segment trees, Dijkstra | | 10^8 | n |