selection

selection provides access to structured spaces of pattern possibility. as part of the structure layer, it enables the generation of symbolic sequences via exhaustive or constrained enumeration, and their refinement via scoring, clustering, or projection into other domains. this domain acts as a curatorial engine - surfacing compact, expressive, and musically relevant material from a large latent pattern space. forms do not emit time-based events, but define symbolic sets and operations over them. results can be forwarded into pitch, onset, density, or other domains for final interpretation. all behavior is deterministic, static, and reproducible. there is no runtime adaptation or feedback.

each form defines a core operation over a symbolic sequence space. all forms accept two parameters a, b ∈ [0,1], remapped to relevant controls.

• subset enumeration

  • behavior: generates all subsequences (or permutations) of a base set, up to a given complexity.
  • analogy: all possible melodies from a limited alphabet
  • a: max length of generated patterns
  • b: enumeration breadth (sparse → exhaustive)

• feature ranking

  • behavior: assigns scores to sequences by structural metrics (e.g. entropy, smoothness), and sorts them.
  • analogy: rating phrases by balance or repetition
  • a: scoring metric
  • b: inclusion threshold (percentile)

• similarity clustering

  • behavior: groups sequences by structural similarity (motivic contour, distribution shape, etc.)
  • analogy: sorting gestures into style categories
  • a: cluster resolution (coarse → fine)
  • b: selected cluster index

• set projection

  • behavior: maps selected symbolic sequences into another domain (pitch, density, etc.)

  • analogy: casting rhythmic structures as melodic intervals

  • a: target domain

  • b: quantization, value mode

• enumerate subsets

  • a: sequence length bound (e.g. 1 = single event, 1.0 = max)
  • b: enumeration scope (sparse = 0, exhaustive = 1)

• rank by feature

  • a: feature metric (entropy, repetition, interval sum)
  • b: percentile filter (top 100% → top 1%)

• cluster by similarity

  • a: cluster granularity
  • b: selected cluster index

• project set

  • a: output domain (pitch, duration, pressure…)

  • b: mapping mode (e.g. linear, quantized, wrap)

• combinatorial depth: allows dense exploration of symbolic material with finite bounds.
• structural filtering: selects meaningful, non-trivial sequences based on clear criteria.
• curation logic: functions as a symbolic sieve between generative space and final usage.
• determinism and hierarchy: nesting allows sequences to be recombined, reused, or modulated structurally.

• stochastic or reactive grammar generation (handled in generative domains)
• real-time adaptive filtering (excluded by design)
• waveform-level feature analysis (outside symbolic scope)
• performance-conditioned selection (belongs to runtime systems)

selection is a general-purpose symbolic engine for structured pattern discovery and reuse. it empowers the user to navigate large combinatorial spaces with minimal entropy, using controlled parameters to yield musical complexity. together with pattern, mutation, and activation, it forms a crucial part of the system's symbolic logic chain: from latent space to expressive material.