arqera-math
Mathematical foundations for AI governance. 9 algorithms, 14 modules, 74 exports, 27 tuned constants. Pure Python. Zero dependencies.
Built by ARQERA for production AI systems that need trust scoring, decision analysis, and self-organizing behaviour.
Why arqera-math?
Most AI governance libraries depend on NumPy, SciPy, or TensorFlow. arqera-math is different:
- Zero dependencies. Pure Python stdlib. No binary wheels. No version conflicts. Runs anywhere Python runs.
- Typed dataclasses everywhere. Every return value is a
@dataclasswith full type annotations. No raw dicts. - Deterministic. No randomness without explicit seeds. Every function is reproducible.
- Self-contained modules. Each module covers one mathematical domain. No circular imports.
- Tested. 105 tests covering 100% of the public API.
Install
pip install arqera-math
Requires Python 3.11+. No dependencies to install -- it is pure stdlib Python.
Quick Start
Bayesian Trust Scoring
Track trust in an entity using Beta-binomial conjugate priors. Each piece of evidence updates the posterior.
from arqera_math import BayesianTrustService
service = BayesianTrustService()
# Create a belief with uniform prior
belief = service.create_belief("agent-1", prior_trust=0.5)
print(f"Initial trust: {belief.mean:.3f}") # 0.500
# Positive evidence increases trust
update = service.update_trust("agent-1", positive_evidence=3)
print(f"After 3 positive: {update.posterior_mean:.3f}") # ~0.615
# Negative evidence decreases trust
update = service.update_trust("agent-1", negative_evidence=2)
print(f"After 2 negative: {update.posterior_mean:.3f}") # ~0.533
# Credible interval
lower, upper = service.compute_credible_interval(belief.alpha, belief.beta)
print(f"95% CI: [{lower:.3f}, {upper:.3f}]")
# Or use the standalone utility for quick calculations
from arqera_math import trust_from_evidence
score = trust_from_evidence(positive=8, negative=2)
print(f"Trust from evidence: {score:.3f}") # ~0.650
PageRank and Graph Analysis
Compute node importance in directed graphs using iterative PageRank.
from arqera_math import simple_pagerank
# Adjacency list: node -> list of nodes it points to
graph = {
"A": ["B", "C"],
"B": ["C"],
"C": ["A"],
"D": ["C"],
}
ranks = simple_pagerank(graph, damping=0.85, iterations=100)
for node, rank in sorted(ranks.items(), key=lambda x: -x[1]):
print(f" {node}: {rank:.4f}")
Information Theory
Measure uncertainty, divergence, and information content.
from arqera_math import entropy, kl_divergence
# Shannon entropy of a probability distribution
h = entropy([0.5, 0.3, 0.2])
print(f"Shannon entropy: {h:.4f} bits")
# KL divergence between two distributions
d = kl_divergence([0.5, 0.3, 0.2], [0.33, 0.33, 0.34])
print(f"KL divergence: {d:.4f}")
Physarum / Stigmergy Dynamics
Pheromone-based route optimization. Successful paths get reinforced, unused paths decay.
from arqera_math import update_pheromone, pheromone_gradient
# Trail intensities (higher = more reinforced)
trails = {"route-A": 0.8, "route-B": 0.3, "route-C": 0.1}
# Reinforce a successful route: tau(t+1) = (1 - rho) * tau(t) + deposit
trails["route-A"] = update_pheromone(trails["route-A"], deposit=0.2, evaporation_rate=0.05)
print(f"Route A after reinforcement: {trails['route-A']:.3f}") # 0.960
# Get gradient -- sorted strongest to weakest
gradient = pheromone_gradient(trails)
for name, intensity in gradient:
print(f" {name}: {intensity:.3f}")
Decision Theory
Score options across multiple weighted criteria.
from arqera_math import decision_rank
criteria = [
{"name": "accuracy", "weight": 0.5},
{"name": "latency", "weight": 0.3, "minimize": True},
{"name": "cost", "weight": 0.2, "minimize": True},
]
options = [
{"name": "model-A", "scores": {"accuracy": 0.95, "latency": 120, "cost": 50}},
{"name": "model-B", "scores": {"accuracy": 0.88, "latency": 30, "cost": 10}},
{"name": "model-C", "scores": {"accuracy": 0.92, "latency": 60, "cost": 25}},
]
ranked = decision_rank(criteria, options)
for result in ranked:
print(f" #{result['rank']}: {result['option_name']} (score: {result['weighted_score']:.3f})")
Algorithms
| Module | What it does |
|---|---|
| Bayesian Inference | Beta-binomial trust model with Fisher information and Cramer-Rao bounds |
| Graph Analysis | PageRank, degree/betweenness centrality, clustering coefficients, HITS |
| Information Theory | Shannon entropy, Renyi entropy, KL divergence, Beta-KL divergence |
| Decision Theory | Weighted multi-criteria decision matrices with sensitivity analysis |
| Multi-Objective Optimization | Pareto frontier extraction, dominance checking, weighted sum scalarization |
| Stigmergy | Pheromone-based route optimization (ant colony style) |
| Quorum Sensing | Hill function biological threshold curves for collective activation |
| Stability Analysis | Lyapunov functions and convergence verification |
| Temporal Dynamics | Trust forecasting, anomaly detection, linear trend analysis |
Plus: control theory (PID controllers), game theory (resource auctions), queueing theory (M/M/1, M/M/c), preconditions (demographic-informed Bayesian priors), and a constants registry with bounds validation.
All Exports
The library exports 74 public symbols across 14 modules. Everything is importable directly from arqera_math:
import arqera_math
print(len(arqera_math.__all__)) # 74
Bayesian: BayesianTrustService, BeliefState, TrustUpdate, FisherInformationResult, bayesian_update, trust_from_evidence, fisher_information, cramer_rao_bound
Graph Analysis: GraphAnalysisService, CentralityMetrics, NodeImportance, simple_pagerank, find_hubs_and_authorities
Information Theory: InformationTheoryService, EntropyMetrics, NodeEntropy, KLDivergenceResult, RenyiEntropyResult, entropy, binary_entropy, kl_divergence, renyi_entropy, beta_kl_divergence
Decision Theory: DecisionMatrix, DecisionCriterion, DecisionOption, DecisionResult, decision_rank, weighted_score
Multi-Objective: ParetoResult, ParetoPoint, ObjectiveWeight, pareto_frontier, dominates, weighted_sum
Stigmergy: StigmergyService, PheromoneTrail, StigmergyMetrics, update_pheromone, pheromone_gradient
Quorum Sensing: QuorumSensingService, QuorumResponse, hill_function
Stability: StabilityService, StabilityAnalysis, lyapunov_function, check_stability
Temporal Dynamics: TrustForecast, TrendAnalysis, TrendPoint, forecast_trust, detect_anomaly, linear_trend
Control Theory: PIDController, ControllerState, ControlAction, simple_pid_step
Game Theory: ResourceAuction, ResourceClaim, AuctionResult, create_claim
Queueing: QueueingService, AgentQueue, QueueMetrics
Preconditions: PreconditionProfile, compute_domain_priors, compute_entity_prior
Constants: MATH_CONSTANTS, ConstantDomain, MathConstant, get_constant, get_constant_info, list_constants_by_domain, validate_constant_update
Constants
27 tuned mathematical constants across 12 domains, each with bounds validation:
from arqera_math import get_constant, list_constants_by_domain, ConstantDomain
# Get a specific constant
value = get_constant("CONDUCTANCE_REINFORCEMENT")
print(value) # 0.3
# List constants by domain
entropy_constants = list_constants_by_domain(ConstantDomain.ENTROPY)
for c in entropy_constants:
print(f" {c.name} = {c.value} (range: {c.min_value}..{c.max_value})")
Used by ARQERA
arqera-math powers the mathematical engine behind ARQERA, an AI operations platform for enterprise governance, trust scoring, and autonomous decision-making.
Every algorithm in this library is battle-tested in production: Bayesian trust updates run on live agent evaluations, PageRank scores real dependency graphs, and Physarum dynamics route actual AI workloads.
Development
# Clone
git clone https://github.com/Arqera-IO/arqera-math.git
cd arqera-math
# Install with dev dependencies
pip install -e ".[dev]"
# Lint
ruff check src/
# Test
pytest tests/ -v
# Quality gate (must pass before commit)
ruff check src/ && pytest tests/ -v
See CONTRIBUTING.md for full contributor guidelines.
License
Apache-2.0. See LICENSE.
Copyright 2026 ARQERA Ltd.
Metadata
Release files for arqera-math 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| arqera_math-0.1.0.tar.gz | 51.8 kB | Without provenance | Details for arqera_math-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| arqera_math-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for arqera_math-0.1.0-py3-none-any.whl |
Total release size: 93.0 kB
Release files / arqera_math-0.1.0.tar.gz
| Download URL | arqera_math-0.1.0.tar.gz |
|---|---|
| Size | 51.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
0d1a221d3cf0f7740c7b0e9f8dfb443f9c0e632e81c3cc7ceea00504fa4917ac
|
|
BLAKE2b-256 checksum How to use checksums |
b9d2fadf4e0967a81738a824371fdd8070056cb2e5ff3e05e2fdaf2290a87b06
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.12
|
Release files / arqera_math-0.1.0-py3-none-any.whl
| Download URL | arqera_math-0.1.0-py3-none-any.whl |
|---|---|
| Size | 41.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
aeff992911361027683dc08b24bcd0e5b669622e35270878d39bee707bf5f02e
|
|
BLAKE2b-256 checksum How to use checksums |
0e1772aba55ec276a2ca323c1bdb50a2a3f4de060d3dbe6610827f78a526fc00
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.12
|
Security for arqera-math 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.