Exploring quantum computation and hybrid quantum‑classical algorithms with a focus on simulation, algorithm development, and applied quantum machine learning.

Repo: Womanium2025--QuantumGaltonBoard
Implementation of Carney & Varcoe’s Universal Statistical Simulator (arXiv:2202.01735). Demonstrates how quantum circuits can simulate Galton Box–style Monte Carlo problems, relevant to high‑dimensional challenges such as particle transport and quantum systems.

• Modular Quantum Galton Board circuits using Hadamard and RY rotations.
• Monte Carlo framing: Galton Box simulation treated as a sampling problem for PDEs.
• Quantum walk dynamics: Hadamard quantum walk distributions compared against classical binomial statistics.
• Noise modeling with backend‑specific emulators and circuit optimization.
• Verified scaling formula: 2n+2 qubits and quadratic depth growth O(n^2).

• Gaussian distributions matched theory with TVD ≈ 0.01–0.02 for n=1,2.
• Exponential distributions (bias) skewed strongly toward Bin 0.
• Hadamard quantum walk reproduced U‑shaped distribution with TVD < 0.01 in noiseless runs.
• Demonstrated exponential speed‑up: 2^n trajectories simulated with only O(n^2) gates.

Repo: Classiq.aztechacks2024
Solo 48‑hour completion of all Classiq challenge functions and bonus algorithm, emphasizing reversible arithmetic, entanglement preparation, and circuit synthesis.

Implemented all 10 reversible arithmetic and oracle functions:

• inplace_square, inplace_linear, inplace_quadratic, inplace_cubic, inplace_exponential
• discrete_log_oracle, inplace_discrete_logarithm
• equality_oracle, inplace_sum, sum_of_squares

Constructed with Classiq’s symbolic quantum types (QNum, QArray[QBit]).
Verified arithmetic logic and oracle behavior.
Synthesized circuits with resource reporting (e.g., inplace_quadratic depth ≈ 12, gate count ≈ 40).

Constructed a 3‑qubit GHZ state using H and CX gates.
Synthesized circuit: 3 qubits, depth ≈ 3.
Simulator results: 000 and 111 states observed with near‑equal probability.
Parsed counts: {'meas': 7.0}: 1042, {'meas': 0.0}: 1006 from 2048 shots.

• All 10 challenge functions implemented and tested successfully.
• GHZ state prepared and validated with correct superposition behavior.
• Resource usage reported for each synthesized circuit.

Repo: Development-of-Novel-Quantum-Algorithms
Exploration of 1D and 2D Ising models as combinatorial optimization problems, implemented with Classiq’s QAOA framework. Includes circuit synthesis, convergence analysis, and Probabilistic Error Cancellation (PEC) for error mitigation.

• 1D Ising model formulated with Pyomo, mapped to QUBO form.
• 2D Ising model (4×4 lattice) with periodic boundary conditions.
• QAOA (Quantum Approximate Optimization Algorithm) applied with 5 layers and CVaR optimization.
• Circuit resources:
  • 1D: width = 6 qubits, depth ≈ 41, gates = {CX: 60, RZ: 60, RX: 30, H: 6}
  • 2D: width = 16 qubits, depth ≈ 91, gates = {CX: 320, RZ: 240, RX: 80, H: 16}
• Trotterization used to approximate Hamiltonian evolution.
• Error mitigation: PEC applied to noisy runs.

• Convergence iterations:
  • 1D noiseless: 60–65
  • 1D trotterized: 75–80
  • 1D noisy: 55–60
  • 1D mitigated (PEC): ~40
  • 2D noiseless: ~65
  • 2D trotterized: 85–90
  • 2D noisy: 90–95
  • 2D mitigated (PEC): 65–70
• Energy landscapes:
  • 1D ground state cost ≈ –180 with probability ≈ 0.77
  • 2D ground state cost ≈ –960 with probability ≈ 0.70
• PEC reduced effective noise, producing energies close to noiseless baselines.

Repo: QML-for-Conspicuity-Detection-in-Production
Hybrid quantum–classical machine learning applied to industrial visual saliency detection. Combines variational quantum circuits with classical deep learning to improve robustness in defect detection and conspicuity analysis.

• Variational Classifier (VQC):
  • PennyLane implementation on 4 qubits.
  • Learned parity function with accuracy = 1.0 after ~35 iterations.
  • Confusion matrix confirmed perfect classification on training and unseen test data.
• Quanvolutional Neural Network (QCNN):
  • Quantum convolution layer applied to MNIST dataset (2×2 pixel patches → 4 expectation channels).
  • Pre‑processed images fed into classical dense network.
  • Validation accuracy improved to ~60–70% with quantum preprocessing compared to ~50–60% baseline.
  • Confusion matrix analysis highlighted class‑specific improvements.
• Hybrid Integration: PennyLane + PyTorch/TensorFlow pipelines, classical optimizers (Adam, Nesterov Momentum).

• Binary classification (parity): Accuracy = 1.0 on both training and unseen test sets.
• Iris dataset (multi‑class): Validation accuracy >70% after ~60 iterations.
• MNIST quanvolution: Quantum preprocessing improved validation accuracy compared to classical baseline.

Qiskit • Classiq SDK • PennyLane • Quantum Circuit Design • QAOA • Variational Quantum Algorithms • Probabilistic Error Cancellation (PEC)

Python • PyTorch • TensorFlow • Data Visualization

LinkedIn: https://www.linkedin.com/in/yasir-mansour-663a02293
GitHub: https://github.com/qcmp34

Last updated: November 25, 2025