selection

the selection domain constructs and filters sets of candidate sequences based on combinatorial logic and quantitative properties. as part of the structure layer, it supports the generation of pattern sets, the extraction of statistical features, and the deterministic selection or sorting of material for further projection into pitch, onset, density, or structural domains. all operations are fully deterministic, with no runtime randomness or adaptive behavior. this domain provides a principled way to explore large spaces of variation and control structures without sacrificing transparency or reproducibility.

each form uses parameters a and b ∈ [0,1], remapped to meaningful musical controls. forms are irreducible, additive, and operate purely on data generation and filtering:

• enumerate subsets

  • behavior: generates all valid subsequences or permutations of a base sequence or set.
  • analogy: listing all valid words from a given alphabet
  • a: maximum size of generated sequences
  • b: coverage mode (0 = minimal, 1 = exhaustive)

• rank by feature

  • behavior: scores generated sequences using a statistical or structural metric, and sorts them.
  • analogy: rating patterns by smoothness or symmetry
  • a: feature type (e.g. entropy, interval sum, symmetry)
  • b: top-k inclusion (0 = broad, 1 = strict)

• cluster by similarity

  • behavior: groups sequences by shared structural traits using deterministic feature space clustering.
  • analogy: sorting textures into perceptual bins
  • a: clustering granularity (coarse to fine)
  • b: chosen cluster index (0 = first, 1 = last)

• project set

  • behavior: maps the selected set of sequences into another control domain (e.g. pitch, density).

  • analogy: casting abstract patterns into physical shape

  • a: target domain (pitch, timing, amplitude, etc.)

  • b: value scaling or quantization mode

• enumerate subsets

  • a: sets maximum length of each pattern (e.g. 1 to full-length)
  • b: controls breadth of enumeration (sparse to exhaustive)

• rank by feature

  • a: selects scoring metric (e.g. entropy = 0.0, smoothness = 0.5, repetition = 1.0)
  • b: sets ranking threshold or percentile

• cluster by similarity

  • a: resolution of clustering (number of groups)
  • b: selects which cluster to use

• project set

  • a: selects target control domain

  • b: transforms value range or domain-specific mapping

• combinatorial depth: enables exhaustive yet structured enumeration of pattern spaces.
• statistical intelligence: allows sorting and grouping not just by form, but by perceptual and structural relevance.
• orthogonality: each form serves a distinct role-generation, analysis, grouping, and projection.
• parametric clarity: all controls are direct, deterministic, and musically meaningful.
• extensibility: new feature extractors or clustering metrics can be added without changing core structure.

• reactive or real-time adaptation: no dynamic selection or response to ongoing output.
• generative grammar models: recursive or stochastic unfolding belongs in the generative pitch or pattern domains.
• mutating sequence behavior: handled by the mutation domain.
• low-level signal patterning: this operates at the symbolic/structural level, not at waveform or modulation resolution.

the selection domain introduces a new layer of control over combinatorial and statistical pattern space. by enumerating, scoring, grouping, and projecting candidate sequences, it enables precise and expressive use of pattern complexity - without sacrificing determinism. as both a generator and curator of structure, selection forms a crucial link between raw material and musically relevant control data across the 声音 system.