Changelog

All notable changes to ATLAS-Q are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.7.0 - 2025-12-16 (IR v1.1.0)

IR Integration - COMPLETE

This release completes the Informational Relativity (IR) integration with pre-computation regime diagnosis.

Pre-Computation Diagnosis (NEW)

• Regime Analyzer (ir_enhanced/regime_analyzer.py): Diagnose quantum problems BEFORE running

  • IR (Informational Relativity) regime: High coherence, quantum advantage likely

  • Transition regime: Marginal coherence, needs careful tuning

  • AIR (Anti-IR) regime: Low coherence, classical methods may be better

  • GO/NO-GO classification using physics-derived e^-2 threshold (0.135)

  • analyze_state_regime(), analyze_hamiltonian_regime(), analyze_mps_bond_regime()

  • predict_quantum_advantage() for resource planning

  • should_use_ir_grouping() for adaptive algorithm selection

Spectral Lifting (NEW)

• Full M_ij Relational Matrix Analysis (ir_enhanced/spectral_lifting.py):

  • compute_relational_matrix() - Build coherence correlation matrices

  • extract_structure_modes() - Find dominant eigenmodes

  • spectral_grouping() - Group observables by spectral similarity

  • coherent_structure_score() - Quantify structure detection quality

Performance Improvements (VALIDATED)

Feature	Improvement	Status
VQE Grouping	4× circuit reduction	Production
Period Finding	42% shot reduction	Production
Variance Reduction	2-60× (VQE)	Production
MPS Scalability	100 qubits in 1.56s	Validated
Memory Compression	10^25× (100 qubits)	Validated

New IR Modules

• gradient_grouping.py - Parameter shift optimization with coherence

• qaoa_grouping.py - Edge-based cost operator grouping

• shadow_tomography.py - Adaptive classical shadows with IR

• state_tomography.py - IR-enhanced state reconstruction

• tdvp_observables.py - Real-time coherence tracking for TDVP

Added

GPU CUDA Backend

• Direct CUDA Driver API integration via ctypes

• Pre-compiled PTX kernels (version-independent)

• Supports single-qubit gates: H, X, Y, Z, RX, RY, RZ

• Supports two-qubit gates: CNOT, CZ, SWAP

• f64 precision for numerical stability

• 2-13× faster than CPU for 15+ qubits

Triton IR Coherence Kernels

• GPU-accelerated coherence computation

• compute_response_coherence_triton() for R-bar, V_phi metrics

• GO/NO-GO classification at GPU speed

Changed

• Improved exception handling in GPU backend

• Updated ir_enhanced/__init__.py exports for regime analyzer

• Coherence-aware VQE now includes pre-computation regime analysis

Fixed

• Fixed Triton IR Coherence benchmark (amplitudes vs responses/phases)

• Fixed UCCSD Ansatz benchmark imports

• Fixed MANIFEST.in to include compiled Rust extensions

Benchmark Results

• 30/31 benchmarks passing (cuQuantum optional skip)

• 7/7 IR Enhanced tests passing

0.6.x Series

0.6.4 - 2025-11-04

Added

UCCSD Ansatz for Molecular VQE

• UCCSD (Unitary Coupled-Cluster Singles and Doubles) (ansatz_uccsd.py): Chemistry-aware variational ansatz

  • OpenFermion integration for fermionic operator generation

  • MPS-compatible implementation (no exponential memory)

  • Pauli string decomposition with apply_pauli_exp_to_mps()

  • Hartree-Fock reference state initialization

  • Compatible with VQE for ground state chemistry calculations

Quantum Chemistry & Optimization Hamiltonians

• Molecular Hamiltonian Builder (mpo_ops.py): PySCF integration for quantum chemistry

  • molecular_hamiltonian_from_specs() - Build electronic structure Hamiltonians

  • Support for H2, LiH, H2O, and custom geometry strings

  • Jordan-Wigner fermion-to-qubit transformation

  • Compatible with VQE for ground state energy calculations

  • 4/4 tests passing in test_molecular_hamiltonians.py

• MaxCut Hamiltonian Builder (mpo_ops.py): QAOA graph optimization

  • maxcut_hamiltonian() - Build MaxCut problem Hamiltonians

  • Weighted and unweighted graph support

  • Automatic edge normalization for undirected graphs

  • Compatible with QAOA for combinatorial optimization

  • 4/4 tests passing in test_maxcut.py

Advanced Tensor Network Features

• Circuit Cutting (circuit_cutting.py): Partition large circuits for simulation

  • Coupling graph analysis and entanglement heatmaps

  • Min-cut and spectral partitioning algorithms

  • Classical stitching with variance reduction

  • 7/7 tests passing in test_circuit_cutting.py

• PEPS (Projected Entangled Pair States) (peps.py): 2D tensor networks

  • True 2D representation for shallow quantum circuits

  • Boundary-MPS contraction strategy

  • PatchPEPS for 4×4 and 5×5 grids

  • Single and two-site gate application

  • 10/10 tests passing in test_peps.py

• Distributed MPS (distributed_mps.py): Multi-GPU scaling

  • Bond-wise domain decomposition across GPUs

  • Overlapped communication and computation

  • Checkpoint/restart for long simulations

  • 10/10 tests passing in test_distributed_mps.py

• cuQuantum Backend (cuquantum_backend.py): Optional NVIDIA acceleration

  • cuTensorNet 25.x integration for tensor operations

  • Automatic fallback to PyTorch if unavailable

  • 2-10× speedup on compatible NVIDIA GPUs (requires cuquantum-python)

  • 11/11 tests passing in test_cuquantum.py (tested with cuQuantum 25.09.1)

  • Install: pip install cuquantum-python (optional, ~320MB)

Changed

Improved Import System (Better UX)

• Direct module imports now supported (__init__.py): Pythonic import pattern

  • New (recommended): from atlas_q import mpo_ops, tdvp, vqe_qaoa

  • Legacy (still works): atlas_q.get_mpo_ops() returns dict

  • Enables IDE autocomplete and type hints

  • Matches standard Python package conventions (like NumPy, PyTorch)

  • Backwards compatible - old getter pattern still supported

0.5.0 - 2025-10-26

Added

GPU Acceleration & Triton Kernels

• Custom Triton kernels for MPS gate operations (1.5-3× speedup)

• GPU-optimized tensor contractions with cuBLAS tensor cores

• Modular exponentiation kernels for period-finding

• 77,000+ ops/sec gate throughput achieved

Tensor Network Features

• Noise Models (noise_models.py): Full Kraus operator framework for NISQ simulation

  • Depolarizing, dephasing, amplitude damping, Pauli noise channels

  • Stochastic noise applicator with reproducible seeds

• Stabilizer Backend (stabilizer_backend.py): Clifford circuit fast path

  • O(n²) complexity via Gottesman-Knill theorem

  • 20× speedup over generic MPS

  • Automatic handoff to MPS for non-Clifford gates

• MPO Operations (mpo_ops.py): Hamiltonian and observable framework

  • Pre-built Hamiltonians: Ising, Heisenberg, custom spin chains

  • Expectation values and correlation functions

• TDVP (tdvp.py): Time evolution for Hamiltonian dynamics

  • 1-site (conserves bond dimension) and 2-site (adaptive) variants

  • Krylov subspace methods for efficient evolution

• VQE/QAOA (vqe_qaoa.py): Variational quantum algorithms

  • Ground state finding and combinatorial optimization

  • Hardware-efficient ansätze with classical optimizer integration

Adaptive MPS Implementation

• Energy-based adaptive truncation with error bounds

• Per-bond dimension caps and global memory budgets

• Mixed precision support (complex64/complex128) with auto-promotion

• Comprehensive statistics tracking and diagnostics

• Canonical forms (left, right, mixed) with robust QR/SVD

Documentation & Testing

• Complete whitepaper documenting architecture and algorithms

• Research paper with mathematical foundations

• 75+ unit tests across unit, integration, and performance suites

• All 7/7 benchmark suites passing

Performance

• 626,454× memory compression (30 qubits)

• 20.4× Clifford circuit speedup

• GPU-accelerated operations throughout

• Demonstrated capacity: 100,000+ qubits (χ=64, moderate entanglement)

Changed

• Updated lazy import structure in __init__.py

• Enhanced numerical stability with multi-driver SVD fallback

• Improved error tracking and diagnostics

Legend

• Added: New features

• Changed: Changes in existing functionality

• Fixed: Bug fixes

• Performance: Performance improvements