Shunyaya Symbolic Mathematics — Function library — quick reference tables (2.49)

Abstract
Compact, copy-paste quick reference for SSM function lifts. Each entry states the magnitude rule (ASCII), domain/range, monotonicity, alignment policy, key guards, and a stable numeric formula where applicable. Under collapse phi(m,a) = m, all rows reduce to their classical counterparts.


Legend (columns)
Name | Lift on (m,a) | Domain → Range | Monotonicity | Alignment | Guards | Stable numerics


Elementary & rational

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
Power k (integer)( m^k , a )R → Rinc on m≥0 if k≥1; even k nonmonotonecarrydeclare branchwise for even k
Root r (principal)( m^(1/r) , a )m≥0 → m≥0increasingcarryerror if m<0scale before pow for large m
Reciprocal( 1/m , a )m≠0 → Rdecreasing on (0,∞)carryerror at m=0fuse as 1/(scaled m)
Rational p/q( p(m)/q(m) , a )q(m)≠0piecewisecarrypoles where q=0use Horner; guard near poles

Inverse trigonometric (principal)

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
asin( asin(m) , a )[-1,1] → [-pi/2, +pi/2]increasingcarryerror ifm
acos( acos(m) , a )[-1,1] → [0, pi]decreasingcarryerror ifm
atan( atan(m) , a )R → (-pi/2, +pi/2)increasingcarryatan via atan2(m,1)
atan2 (y,x)( atan2(y,x) , a’ )(x,y)≠(0,0) → (-pi,+pi]quadrant-awarea’ from weightserror at originuse builtin atan2

Hyperbolic & inverse hyperbolic

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
sinh( 0.5*(e^m – e^-m) , a )R → Rincreasingcarryexp-splitting; one-sided scaling for
cosh( 0.5*(e^m + e^-m) , a )R → [1,∞)dec on (-∞,0], inc on [0,∞)carryexp-splitting; branchwise order
coth( cosh(m)/sinh(m) , a )R{0} → (-∞,-1)∪(1,∞)decreasing on each sidecarryundefined at 0use 1/tanh(m)
asinh( log(m + sqrt(m^2+1)) , a )R → Rincreasingcarryfor
acosh( log(m + sqrt(m-1)*sqrt(m+1)) , a )m≥1 → [0,∞)increasingcarryerror if m<1for m≫1 use log(2*m)
acoth( 0.5*log((m+1)/(m-1)) , a )m>1 → (-∞,0)∪(0,∞)odd; sign-monotonecarry

Piecewise, indicator, rounding

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
sgn( -1, +1 ) / (0,+1) / ( +1,+1 )R → {-1,0,1}stepconstants +1tie policy at 0
Heaviside H( 0,+1 ) / (1/2,+1) / (1,+1)R → {0,1/2,1}stepconstants +1choose H(0)
ramp / ReLU( m,a ) if m>0 else (0,+1)R → [0,∞)nondecreasingcarry or +1 at 0tie policy at 0
clip [L,U](L,+1)/(m,a)/(U,+1)L≤Unondecreasingcarry-from-sourceswap or error if L>U
floor( floor(m) , a )R → Zstepcarry
ceil( ceil(m) , a )R → Zstepcarry
trunc( trunc(m) , a )R → Zstepcarry
round( round(m) , a )R → Zstepcarryrounding_modeulp-aware half detection

Selectors & order

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
min_betaselect by S_betaR^2 → Rlatticecarry from selectedtie rule
max_betaselect by S_betaR^2 → Rlatticecarry from selectedtie rule
clamp_betamin_beta(max_beta(x,L),U)L≤Uboundedfrom selectedswap or error

S_beta(m,a) = m*(1 – beta*(1 – a)), beta in [0,1].


Smooth nonlinearities

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
sigmoid( 1/(1+e^-m) , a )R → (0,1)increasingcarry0.5*(1+tanh(m/2))
softplus( log(1+e^m) , a )R → (0,∞)increasingcarrypiecewise: m+log1p(e^-m) / log1p(e^m)
softsign( m/(1+m) , a )R → (-1,1)increasingcarry

Series, analytic, transforms

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
Power series( Σ c_n (m-m0)^n , a )m-m0<Ras fcarry
exp( Σ m^n/n! , a )entireincreasingcarrytruncation with remainder bounds
log(1+m)( Σ (-1)^{n+1} m^n/n , a )m<1increasingcarry
Z-transform( Σ m[n] z^{-n} , a_Z )classical ROCrapidity meanROC guard
Laplace( ∫ m(t) e^{-s t} dt , a_L )Re(s)>sigma0rapidity meanROC guard

Alignment summary (transforms): a_T = tanh( Σ w u / Σ w ), with u=atanh(a), weights typically |m|^gamma.


Special functions

NameLift on (m,a)Domain → RangeMonotonicityAlignmentGuardsStable numerics
Gamma( Gamma(m) , a )m>0 → (0,∞)unimodal (min near 1.4616)carryerror if m≤0 (default)use lgamma; Stirling for large m
Beta( B(x,y) , a_B )x>0,y>0decreasing in each argrapidity mean over inputsdomain guarduse betaln then exp if safe
erf / erfc( erf(m)/erfc(m) , a )Rinc / deccarryseries small
Bessel J,Y( J_nu(m), Y_nu(m) , a )m≥0 / m>0osc.carryY singular at 0scaled forms, recurrences
Bessel I,K( I_nu(m), K_nu(m) , a )m>0I inc, K deccarrysingular at 0scaled IK to avoid overflow
Polylog Li_s( Li_s(m) , a )m<1 (principal)carry
Lambert W_0( W_0(m) , a )m≥-1/eincreasingcarryerror outsideseries small; log-log for large m
Lambert W_-1( W_-1(m) , a )-1/e≤m<0decreasingcarrybranch selectasymptotic near 0-

Multivariate, inverse, inequalities

NameLift on inputsDomain → RangeMonotonicityAlignmentGuardsStable numerics
Aggregator f*( f(m_i) , a’ )per fper frapidity mean with w_i=∂f/∂m_i*
Inverse of monotone f( f^{-1}(y) , a_y )y in rangeinc/dec mapscarry from targetbranch select if nonmonotoneNewton with guards
Implicit F=0( g(p) , a’ )IFT regionas grapidity mean with s_i=∂g/∂p_i
Inequalitiesf*(x)piecewisepreserves/reverses on magnitudesalignment metadata onlydeclare domains

Vectors, matrices, convolution

NameLiftDomain → RangeMonotonicityAlignmentGuardsStable numerics
A+Bentrywisematriceslinearrapidity-weighted sum per entryweights
c*Aentrywisematriceslinearconstants +1
A*Bproductshapes matchbilinearper-term M2 then weighted mean per entryweights
tr(A),A_Fscalarany
det(A)scalarnonsingularpolycofactor-weighted meanwarn near singularadjoint weights
exp(A), log(A)spectrallog on SPDper eigencurveeigen-channel aggregationdomain guardsscaling & squaring
Convolutionclassical on mmode/causal as setlinearrapidity mean withk*
Correlationclassical on mbilinearper-term M2 then mean

Takeaway
Use the rows above as a practical index: apply the classical magnitude formula, carry or aggregate alignment exactly as stated, enforce the guard, and prefer the stable numeric identity. Collapse (a=+1) reproduces the standard function.


Navigation
Previous → Convolution & correlation (scalar kernels) (2.48)
Next → Manifest templates & examples (2.50)


Disclaimer
Observation only. Results reproduce mathematically; domain claims require independent peer review. Defaults: mult_mode = M2 for products, rapidity-weighted sums for additions/aggregations, clamp_eps = 1e-6, |a| < 1. All formulas are presented in plain text. Collapse uses phi(m,a) = m.