Skip to main content

quantum-hamiltonian-learning-long-times

Hamiltonian learning methodology from single time evolution at arbitrarily long times. Covers local Hamiltonian families, normalization conditions, and probabilistic learning guarantees.

الانتقال إلى التثبيت

معلومات المصدر

المستودع
hiyenwong/ai_collection
آخر نشاط في المصدر
٨ يونيو ٢٠٢٦ في ٠٨:١١
لغة SKILL.md المكتشفة
الإنجليزية
النجوم
٢
التفرعات
٠

خيارات التثبيت

يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.

مراجعة ملفات المصدر

اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.

عرض SKILL.md

SKILL.md
تعليمات المصدر · معاينة للقراءة فقط
name
quantum-hamiltonian-learning-long-times
category
quantum-computing
description
Hamiltonian learning methodology from single time evolution at arbitrarily long times. Covers local Hamiltonian families, normalization conditions, and probabilistic learning guarantees.
activation
learning hamiltonians long times, hamiltonian learning, quantum learning theory, local hamiltonian, time evolution learning, quantum system identification, 哈密顿量学习
# Quantum Hamiltonian Learning at Long Times ## Description Methodology for learning unknown n-qubit Hamiltonians from single time evolution U = e^{-iHt} where t may be arbitrarily large. Provides provable results for broad families of local Hamiltonians, overcoming the challenge that long-time evolution obscures individual terms. Based on arXiv:2606.05690 (Cedillo, Cotler, Huang). ## Core Problem Standard Hamiltonian learning fails at large times because e^{-iHt} accumulates global phase information that masks individual coupling terms. This paper shows that for **random local Hamiltonians**, learning is still possible with high probability over H and t. ## Mathematical Framework ### Hamiltonian Learning Setup - **Input**: Single unitary U = e^{-iHt} for unknown H, at possibly large t - **Goal**: Recover H (or key properties of H) from U - **Challenge**: At large t, eigenvalues wrap around the unit circle (phase 2π ambiguity) ### Key Results 1. **Probabilistic learnability**: For broad families of local Hamiltonians, with high probability over random H and t, the Hamiltonian is learnable 2. **Normalization constraint**: Any sum of local observables A that is normalized and satisfies [A, H] = 0 must be trivial 3. **Local Hamiltonian families**: Results apply to geometrically local, k-local, and random Hamiltonians ### Learning Algorithm Structure ``` Given: U = e^{-iHt}, time t (arbitrary) 1. Decompose H = Σ_j h_j P_j (Pauli basis) 2. For random H, show that eigenvalue spacing prevents destructive interference 3. Use spectral properties to identify coupling terms 4. Probabilistic guarantees over H and t distributions ``` ## Usage Patterns ### Pattern 1: Quantum System Identification When given access to a quantum system's time evolution: 1. Collect measurement data at different observables 2. Use the probabilistic learning framework to identify Hamiltonian terms 3. Validate against known physical constraints (locality, symmetry) ### Pattern 2: Long-Time Dynamics Analysis When studying quantum systems at large evolution times: 1. Account for eigenvalue wrapping (2π phase ambiguity) 2. Use random Hamiltonian assumptions to break degeneracies 3. Apply normalization conditions to filter spurious solutions ### Pattern 3: Quantum Machine Learning For quantum ML tasks involving Hamiltonian parameter estimation: 1. Frame as learning problem: recover parameters from e^{-iH(θ)t} 2. Leverage the paper's results to design learning algorithms 3. Use the probabilistic guarantees for convergence analysis ## Key Concepts - **Local Hamiltonians**: H = Σ_i h_i where each h_i acts on O(1) qubits - **Normalization**: ||A||_F = 1 for observable A - **Commutation condition**: [A, H] = 0 implies A is trivial (for random H) - **Probabilistic framework**: Results hold with high probability over H and t ## Applications - Quantum system characterization and benchmarking - Quantum error correction (learning noise Hamiltonians) - Quantum simulation validation - NISQ device calibration - Quantum machine learning (Hamiltonian parameter estimation) ## Error Handling ### Phase Ambiguity at Long Times - Problem: eigenvalues wrap around 2π, causing term cancellation - Solution: Use probabilistic analysis over H and t distributions to avoid worst-case configurations ### Non-Local Hamiltonians - Problem: Results only proven for local Hamiltonian families - Solution: Verify locality structure before applying; for non-local H, use alternative learning protocols ## Related Skills - quantum-learning-theory - quantum-system-identification ## Resources - arXiv: 2606.05690 - "Learning Hamiltonians at Long Times" (Cedillo, Cotler, Huang)
عرض على GitHub