At1uri's picture
Upload folder using huggingface_hub
be4df46 verified
|
Raw History Blame Contribute Delete
5.57 kB
metadata
title: Quantum Portfolio Optimization Studio
emoji: ⚛️
colorFrom: purple
colorTo: blue
sdk: static
pinned: false
tags:
  - quantum-computing
  - qaoa
  - vqe
  - ising-hamiltonian
  - qubo
  - portfolio-optimization
  - quantitative-finance
short_description: Quantum VQE and QAOA Ising Hamiltonian portfolio optimizer.

Quantum Portfolio Optimization Studio

An interactive, client-side, research-grade quantum computational engine for solving NP-hard cardinality-constrained portfolio optimization and combinatorial arbitrage problems via Variational Quantum Eigensolver (VQE) and Quantum Approximate Optimization Algorithm (QAOA).

1. Theoretical Formulation & Hamiltonian Mapping

Markowitz mean-variance optimization with cardinality constraints and transaction penalties is mapped to a Quadratic Unconstrained Binary Optimization (QUBO) problem:

min⁡x∈{0,1}nC(x)=q xTΣx−μTx+λbudget(∑i=1nxi−K)2\min_{\mathbf{x} \in \{0, 1\}^n} \mathcal{C}(\mathbf{x}) = q \, \mathbf{x}^T \boldsymbol{\Sigma} \mathbf{x} - \boldsymbol{\mu}^T \mathbf{x} + \lambda_{\text{budget}} \left(\sum_{i=1}^n x_i - K\right)^2

where:

  • $\mathbf{x} \in {0, 1}^n$ is the binary asset selection vector,
  • $\boldsymbol{\Sigma} \in \mathbb{R}^{n \times n}$ is the asset return covariance matrix,
  • $\boldsymbol{\mu} \in \mathbb{R}^n$ is the expected asset return vector,
  • $q \ge 0$ is the investor risk-aversion coefficient,
  • $K \in {1, \dots, n}$ is the target cardinality (number of selected assets),
  • $\lambda_{\text{budget}} > 0$ is the quadratic penalty enforcing the exact cardinality constraint.

Mapping to the Ising Spin Glass Model

Using the Jordan-Wigner / spin-1/2 mapping $x_i = \frac{I - \sigma_i^z}{2}$ where $\sigma_i^z = \begin{pmatrix} 1 & 0 \ 0 & -1 \end{pmatrix}$ is the Pauli-$Z$ operator:

HC=∑i=1nhiσiz+∑1≤i<j≤nJijσizσjz+Coffset⋅I\mathcal{H}_C = \sum_{i=1}^n h_i \sigma_i^z + \sum_{1 \le i < j \le n} J_{ij} \sigma_i^z \sigma_j^z + C_{\text{offset}} \cdot I

The local longitudinal magnetic field $h_i$ and the exchange coupling $J_{ij}$ are derived analytically:

hi=−12μi+q∑k≠iΣik+q2Σii+λbudget(1−K)h_i = -\frac{1}{2} \mu_i + q \sum_{k \neq i} \Sigma_{ik} + \frac{q}{2} \Sigma_{ii} + \lambda_{\text{budget}} (1 - K)

Jij=q4Σij+λbudget2J_{ij} = \frac{q}{4} \Sigma_{ij} + \frac{\lambda_{\text{budget}}}{2}

Coffset=∑i(q4Σii−12μi)+∑i<j(q4Σij+λbudget2)+λbudgetK2C_{\text{offset}} = \sum_{i} \left(\frac{q}{4}\Sigma_{ii} - \frac{1}{2}\mu_i\right) + \sum_{i < j} \left(\frac{q}{4}\Sigma_{ij} + \frac{\lambda_{\text{budget}}}{2}\right) + \lambda_{\text{budget}} K^2


2. Quantum Algorithms

A. Quantum Approximate Optimization Algorithm (QAOA)

QAOA initializes $n$ qubits in an equal superposition state:

∣+⟩⊗n=12n∑x∈{0,1}n∣x⟩|+\rangle^{\otimes n} = \frac{1}{\sqrt{2^n}} \sum_{x \in \{0, 1\}^n} |x\rangle

Applying $p$ alternating unitary layers parameterized by angles $(\boldsymbol{\gamma}, \boldsymbol{\beta}) \in \mathbb{R}^{2p}$:

∣γ,β⟩=∏l=1pUB(βl) UC(γl) ∣+⟩⊗n|\boldsymbol{\gamma}, \boldsymbol{\beta}\rangle = \prod_{l=1}^p U_B(\beta_l) \, U_C(\gamma_l) \, |+\rangle^{\otimes n}

where:

  • Cost Unitary: $U_C(\gamma_l) = e^{-i \gamma_l \mathcal{H}C} = \prod_i e^{-i \gamma_l h_i \sigma_i^z} \prod{i < j} e^{-i \gamma_l J_{ij} \sigma_i^z \sigma_j^z}$
  • Mixer Unitary: $U_B(\beta_l) = e^{-i \beta_l \mathcal{H}M} = \prod{i=1}^n e^{-i \beta_l \sigma_i^x} = \prod_{i=1}^n R_x(2\beta_l)_i$

The expectation value is minimized via a classical-quantum hybrid loop:

min⁡γ,β⟨HC⟩=⟨γ,β∣HC∣γ,β⟩\min_{\boldsymbol{\gamma}, \boldsymbol{\beta}} \langle \mathcal{H}_C \rangle = \langle \boldsymbol{\gamma}, \boldsymbol{\beta} | \mathcal{H}_C | \boldsymbol{\gamma}, \boldsymbol{\beta} \rangle

B. Variational Quantum Eigensolver (VQE)

VQE prepares a parameterized ansatz $|\psi(\boldsymbol{\theta})\rangle$ using a hardware-efficient circuit:

  1. Layer of single-qubit rotations $R_y(\theta_{i, 0}) = \exp(-i \frac{\theta_{i,0}}{2} \sigma_i^y)$.
  2. Entanglement layer using CNOT or Controlled-Z ($CZ$) gates across linear or full topologies.
  3. Repetition for depth $d \ge 1$.

By the Rayleigh-Ritz variational principle:

⟨HC⟩θ=⟨ψ(θ)∣HC∣ψ(θ)⟩⟨ψ(θ)∣ψ(θ)⟩≥E0\langle \mathcal{H}_C \rangle_{\boldsymbol{\theta}} = \frac{\langle \psi(\boldsymbol{\theta}) | \mathcal{H}_C | \psi(\boldsymbol{\theta}) \rangle}{\langle \psi(\boldsymbol{\theta}) | \psi(\boldsymbol{\theta}) \rangle} \ge E_0

where $E_0$ is the exact ground-state energy corresponding to the globally optimal portfolio bitstring $x^*$.


3. Quantum State Diagnostics

  • Bipartite Von Neumann Entanglement Entropy: $$S(\rho_A) = -\text{Tr}(\rho_A \log_2 \rho_A) = -\sum_{k} \lambda_k \log_2 \lambda_k$$ measures non-classical correlations synthesized by entangling gates.
  • Single-Qubit Bloch Vector Coordinates: $$\mathbf{r}_i = (\langle \sigma_i^x \rangle, \langle \sigma_i^y \rangle, \langle \sigma_i^z \rangle) = (\text{Tr}(\rho_i \sigma^x), \text{Tr}(\rho_i \sigma^y), \text{Tr}(\rho_i \sigma^z))$$
  • Approximation Ratio: $$\alpha = \frac{\langle \mathcal{H}C \rangle - E{\max}}{E_0 - E_{\max}} \in [0, 1]$$

4. Architecture & Capabilities

  • Zero-Latency In-Browser Simulator: Full $2^n$-dimensional complex statevector $\in \mathbb{C}^{2^n}$ matrix-vector algebra implemented directly in client-side JavaScript.
  • Exact Classical Ground State Reference: Brute-force evaluates all $2^n$ basis states to compute the global optimum and verify quantum convergence.
  • Interactive Multi-Qubit Visualizations:
    • Live energy landscape and variational optimization trajectory.
    • Quantum measurement probability distribution $|\langle x | \psi \rangle|^2$.
    • Bloch Sphere vector coordinates per qubit.
    • Real-time bipartite entanglement entropy curve.
    • Quantum circuit schematic visualizer.