Shunyaya Symbolic Mathematical Symbols – Optional Environment Gate (Operator Finalizer) (1.6)

What this page covers
An optional, bounded gate that attenuates the alignment channel only (never magnitudes). We define a practical lane recipe, list properties (collapse parity, bounds, monotone calming), and provide guards, manifest keys, and QA checks.

Definition (alignment-only attenuator)
Given an operator output (m_op, a_op), apply a gate g_t in [0,1] to alignment only:

a_env = clamp_a( g_t * a_op , eps_a )
m_env = m_op

g_t may be a scalar or a field broadcastable to the shape of a_op.

Example lane recipe (bounded, with memory)

clip(x, lo, hi) = min( max(x, lo), hi )

# inputs (each in [0,1])
Z_t in [0,1]            # lane stress
A_t = 1 / (1 + Z_t)     # instantaneous calm (in [0,1])

# memory accumulator (in [0,1])
Q_t = rho * Q_prev + (1 - rho) * clip( A_t - Z_t , 0 , 1 )

# gate (pre-clamp)
g_t = ( 1 / ( 1 + Z_t + kappa * abs(Z_t - A_t) ) ) * ( 1 - exp(-mu * Q_t) )

# clamp to [0,1]
g_t = clip( g_t , 0 , 1 )

Parameters: rho in [0,1], kappa >= 0, mu >= 0, with published initial Q0 in [0,1].

Properties

  • Collapse parity phi( (m_op, a_env) ) = m_op
  • Bounds (if |a_op| <= 1 - eps_a and g_t in [0,1]) |a_env| <= 1 - eps_a
  • Monotone calming g_t = 1 -> a_env = a_op g_t = 0 -> a_env = 0 # for fixed a_op, |a_env| is nondecreasing in g_t
  • Stateless vs stateful
    Set mu = 0 (or rho = 0) for a memoryless gate. Positive mu and rho produce history-aware calming via Q_t.
  • Composability (pointwise)
    Gate after each operator; repeated gates are safe. With clamping, effective gain is bounded by the product of gains.

Guards and manifest keys

gate_used   in {"on","off"}   # if "off": set g_t = 1 everywhere
lane_recipe : <string>        # e.g., "ZE-lane-v1" (publish the recipe name)
kappa, mu   : floats >= 0
rho         : float in [0,1]
Q0          : float in [0,1]  # initial memory
Z_t_source  : <declared>      # how Z_t is computed

Operational rules

  • Apply the gate after each operator’s alignment is computed; never change magnitudes.
  • Publish that Z_t, A_t, Q_t, g_t are clamped to [0,1].
  • Ensure final clamp: a_env = clamp_a( g_t * a_op , eps_a )
  • For tensors, g_t must be broadcast-compatible with a_op.

Tiny examples (illustrative)

eps_a = 1e-6

a_op = +0.80, g_t = 0.50  ->  a_env = clamp_a( 0.40 , eps_a ) = +0.40
a_op = -0.60, g_t = 0.25  ->  a_env = clamp_a( -0.15, eps_a ) = -0.15

Minimal QA invariants (L0–L1 for gating)

# L0 collapse parity
assert phi( (m_op, clamp_a(g_t*a_op, eps_a)) ) == m_op

# L0 bounds
assert abs( clamp_a(g_t*a_op, eps_a) ) <= 1 - eps_a

# L1 gate OFF equals identity on alignment (within numeric tolerance)
if gate_used == "off":
    assert clamp_a(1*a_op, eps_a) == a_op

# L1 monotonic calming (sampled)
if g1 <= g2:
    assert abs( clamp_a(g1*a_op, eps_a) ) <= abs( clamp_a(g2*a_op, eps_a) )


Navigation
Previous: Collapse-Safety Contract (1.5)
Next: Minimal Manifest (1.7)