Merged
Conversation
Signal params (emission_rate, decay_rates) are now per-creature searchable parameters rather than platform-level constants. Removed EMISSION_RATE=0.02 and _DECAY_RATES from flowlenia.py entirely. params.py / result.py - emission_rate: scalar [0.001, 0.05], sigma 0.005 - decay_rates: (C,) vector [0, 0.9], sigma 0.05 - Both added to SIGNAL_PARAMETER_SPECS, sample_random, mutate, in_range - _SignalParams TypedDict extended with both fields flowlenia.py / localized.py - Params gains emission_rate: float | None and decay_rates: Tensor | None - _decay built from params.decay_rates at construction time - step() uses params.emission_rate (zero-safe fallback when None) - localized.py uses per-species emission_rate for each species rollout.py / run.py - _params_dict_to_tensors and _params_from_creature convert both new params quality.py - CREATURE_MASS_FLOOR raised 0.1 -> 0.2 (stricter mass bleed selection) - signal_retention term: quality = 0.7 * compactness + 0.3 * (final_mass/initial_mass) applied when final_signal_mass > 0 or initial_total > initial_mass Presets / CLI - standard_preset: 300 -> 500 steps - signal_preset(): new, 192x192, 800 steps - CLI auto-selects signal_preset when --signal-field and no --steps override Tests - test_rollout.py: updated standard preset assertion (300->500), added signal_preset test - test_signal.py: new keys in all key-presence checks; emission_rate/decay_rates range tests; test_alive_filter_signal_creature_mass_floor_is_stricter; test_signal_retention_boosts_quality - 389 passed, 0 pyright errors
SimOutput captures signal dynamics during simulation: - signal_total_history: total signal mass per step (parallel to mass_history) - signal_channel_snapshots: (n_snapshots, C) mean per channel at snapshot steps - species_signal_received: (S, n_snapshots, C) mean signal over each species territory at snapshot steps (hetero only) analytics.py: two new functions - compute_signal_observables(): mass fraction, dominant channel history - compute_signal_observables_hetero(): receptor alignment and emission-reception compatibility matrix EcosystemMeasures: 6 new fields, all serialized to summary.json build_index.py: reads and passes all new fields to ecosystem template smoke_ecosystem.sh: verifies all signal fields present in summary.json 8 new tests in test_analytics.py 397 passed, 0 errors
ecosystem.html - "total mass conserved" label replaced with context-aware message distinguishing signal-active runs from true mass-conserved runs - SIGNAL badge: teal bordered badge appears in run header when has_signal_data is true, with tooltip explaining signal-enabled archives - Outcome label tooltips: hover over any outcome badge shows plain-English description of what that label means (coexistence, exclusion, merger, fragmentation, cannibalism, stable_isolation, partial_clustering) - Signal charts section (gated on has_signal_data): signal total history, signal mass fraction [0,1], receptor alignment per species per snapshot, emission-reception compatibility heatmap (S×S) - Signal overlay checkbox on GIF viewport: toggles a teal overlay showing signal field active state with dominant channel and magnitude readout archive.html - SIGNAL badge in run header (reads has_signal from manifest.json) - buildParamsTable extended: when params contain signal keys, renders a SIGNAL FIELD section with emission_rate scalar, per-channel bar charts for emission_vector (teal), receptor_profile (purple, symmetric ±1 range), decay_rates (amber), and signal kernel radius render.py - render_archive_page gains has_signal: bool = False parameter, passed to archive.html template build_index.py - _render_run reads signal_field from manifest.json, passes has_signal to render_archive_page - outcome_tooltips dict passed to ecosystem template - has_signal_data detection from signal_total_history index.html (System tab) - Signal field description updated: lists all 8 searchable signal params (emission_vector, receptor_profile, emission_rate, decay_rates, signal_kernel_*), mentions 800-step auto-selection, retention quality term README - v3.4.0 roadmap entry ticked with full feature list - Test count: 397 tests, 0 warnings
…tem tab restructure
quality.py -- fully rewritten
Three-component score replaces single-point compactness:
non-signal: q = 0.6 × compactness + 0.4 × stability
signal: q = 0.5 × compactness + 0.3 × stability + 0.2 × retention
compactness = min(compact(state_T/2), compact(state_T))
Two-point minimum catches early-peak degraders that score 0.99 at T
but were already diffuse at T/2. Key insight: almost all viable solitons
score >0.95 at the final step, making single-point compactness nearly
constant across the population. The midpoint term is where selection
pressure actually comes from.
stability = clip(1 - drift / 0.2, 0, 1)
Continuous version of the persistent filter. Borderline survivors
(drift=0.19) score near 0; rock-stable creatures (drift=0.01) score ~1.
retention = clip(final_mass / initial_mass, 0, 1)
Signal runs only. Penalises runaway emission.
descriptors.py
RolloutTrace gains midpoint_state: np.ndarray | None = None
rollout.py
Midpoint state captured at config.steps // 2, passed to RolloutTrace
test_quality.py -- 4 new tests
midpoint degrader detection, stability term direction, unit interval,
None midpoint fallback
docs/signal-field.svg -- new diagram
One-step signal field mechanics in the existing visual language:
mass↔signal exchange via emission (amber) and reception (purple),
decay (red), conservation note, parameter labels
index.html
System tab restructured with internal anchor mini-nav:
Architecture · Search · Quality metric · Ecosystem · Signal field · Descriptors
New Quality metric section: prose + math blocks with the full formula
New Signal field section: parameter table in monospace math, step-by-step
physics prose, diagram embed, observable descriptions
Signal prose removed from Ecosystem section (now its own section)
build_index.py
signal-field.svg loaded and passed to index template as svg_signal_field
README
Quality metric section updated with formula code block
Signal field section gains diagram reference
Test count: 401
SimOutput gains signal_sum_snapshots: (n_snapshots, H//4, W//4) float32 signal.sum(dim=-1) downsampled 4x, captured at every snapshot step in both homo and hetero loops. ~10KB per snapshot vs 2.3MB for full field. _colorize_signal_frame(): teal colormap for homo, species-colored for hetero using the same SPECIES_PALETTE blended by downsampled ownership weights. signal.gif generated alongside ecosystem.gif in _compute_outputs, upsampled 4x back to match mass GIF dimensions. build_index.py: signal.gif loaded as signal_gif_src, passed to template. ecosystem.html: Mass/Signal tab toggle top-left of GIF viewport -- swaps ecoGif / ecoGifSignal visibility. Both GIFs are embedded; toggle is instant, no re-fetch. Signal overlay checkbox and canvas removed entirely (was decorative noise). Mass chart label: "mass field only · signal tracked separately". smoke_ecosystem.sh: signal.gif presence asserted for signal-active runs. 401 passed, 0 errors, 0 warnings
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
v3.4.0 — Signal field physics, quality metric overhaul, viewer improvements
This release completes the signal field introduced in v3.3.0 with corrected physics, a new multi-component quality metric, spatial signal visualization, and substantial viewer improvements across the archive and ecosystem pages.
Signal field physics corrections
emission_rate(scalar) anddecay_rates(16-channel vector) are now per-creature searchable parameters replacing hardcoded platform constants--signal-fieldCREATURE_MASS_FLOORraised to 0.2 to create selection pressure against mass bleedQuality metric overhaul
q = w_c · min(compact(T/2), compact(T)) + w_s · stability + w_r · retentionclip(1 − drift/0.2, 0, 1)clip(final_mass/initial_mass, 0, 1)weighted at 0.2midpoint_statecaptured during rollout and stored onRolloutTraceSignal observables
SimOutputcapturessignal_total_history,signal_channel_snapshots,species_signal_received, andsignal_sum_snapshots(downsampled spatial signal for GIF generation)EcosystemMeasuresgains 6 new fields: signal total history, mass fraction, channel snapshots, dominant channel history, receptor alignment, emission-reception compatibility matrixViewer improvements
Infrastructure
pyrightwarning count reduced to 0 by suppressing third-party stub noise categoriessmoke_ecosystem.shverifies signal.gif presence for signal-active runs