Comprehensive Search Methods Catalog

Input: $ARGUMENTS

Overview

An exhaustive inventory of search methods found in nature, science, technology, and human systems. Understanding these methods enables borrowing strategies across domains.

Meta-question: Is intelligence just “how well you can search”?

Steps

Step 1: Identify the Search Problem

1. What are you searching FOR? (solution, information, option, path, configuration)
2. What is the search SPACE? (finite/infinite, discrete/continuous, structured/unstructured)
3. What CONSTRAINTS exist? (time, resources, access, knowledge)
4. What does SUCCESS look like? (exact match, good enough, optimal, novel)

Step 2: Catalog of Search Methods by Domain

Exhaustive / Systematic:

• Brute force enumeration — try everything
• Grid search — sample at regular intervals
• Branch and bound — eliminate provably bad regions
• Backtracking — try, fail, undo, try different path
• Dynamic programming — break into overlapping subproblems

Heuristic / Guided:

• Hill climbing — always move toward improvement
• Simulated annealing — accept worse moves early, fewer later
• Tabu search — remember where you’ve been, don’t repeat
• Beam search — keep top-N candidates, expand all
• A* / best-first — use estimated distance to guide exploration

Evolutionary / Population:

• Genetic algorithms — mutate, crossover, select
• Particle swarm — each agent influenced by personal and group best
• Ant colony — pheromone trails reinforce good paths
• Differential evolution — perturb using differences between candidates

Probabilistic / Statistical:

• Monte Carlo — random sampling
• Bayesian optimization — model the objective, sample where uncertain
• MCMC (Markov Chain Monte Carlo) — random walk biased toward good regions
• Thompson sampling — sample from posterior, act on sample
• Random restart — run local search many times from random starts

Adversarial / Game-Theoretic:

• Minimax — assume opponent plays optimally
• Alpha-beta pruning — skip branches that can’t affect outcome
• Nash equilibrium search — find stable strategy profiles
• Regret minimization — minimize worst-case regret over time

Learning / Adaptive:

• Reinforcement learning — reward good paths, punish bad ones
• Active learning — query the most informative examples
• Curriculum learning — start easy, increase difficulty
• Transfer learning — apply learned search strategy to new domain

Biological / Natural:

• Chemotaxis — follow gradient (bacteria)
• Lévy flights — short local searches + occasional long jumps (foraging)
• Stigmergy — modify environment to guide future search (ants, termites)
• Immune system — generate random antibodies, amplify matches
• Neural search — pattern completion from partial cues (brain)

Social / Human:

• Ask an expert — skip search, use someone else’s result
• Crowdsourcing — distribute search across many agents
• Market mechanisms — price signals aggregate distributed information
• Gossip protocols — spread information through local exchanges
• Institutional memory — search organizational records

Constraint-Based:

• SAT solvers — systematic logical constraint satisfaction
• Linear programming — optimize within linear constraints
• Constraint propagation — eliminate impossible values before searching

Step 3: Match Methods to Problem Characteristics

Step 4: Apply to Input

1. Classify the input’s search problem (from Step 1)
2. Match to appropriate methods (from Step 3)
3. For each recommended method:
   • How would it apply concretely?
   • What resources does it require?
   • What are its failure modes?
4. Recommend primary method + backup

Step 5: Cross-Domain Transfer

1. Is the input’s domain using only its “native” search methods?
2. What methods from OTHER domains might apply?
3. Which cross-domain transfer is most promising?
4. → INVOKE: /cda (cross-domain analogy) if transfer looks productive

Step 6: Report

SEARCH METHOD ANALYSIS:
Problem type: [classification]
Search space: [size, structure, smoothness]
Constraints: [time, resources, knowledge]

Recommended methods:
1. [Primary] — why it fits, how to apply
2. [Backup] — when to switch to this
3. [Cross-domain] — borrowed from [domain], novel application

Methods to AVOID: [which and why]

When to Use

• Stuck on a problem and need a new approach
• Want to systematically consider all search strategies
• Looking for cross-domain inspiration
• Designing an algorithm or process

Verification

• Search problem clearly characterized
• Multiple method categories considered
• Methods matched to problem characteristics
• Cross-domain transfer explored
• Failure modes of recommended methods noted