Repurposed.

.gitignore

@@ -133,3 +133,4 @@ venv.bak/
 .stestr/
 
 # project-specific ignores
+results@0,[2.600974180557, 10.435233760548].png results@1,[2.600974180557, 10.435233760548].png results@2,[2.600974180557, 10.435233760548].png results@3,[2.600974180557, 10.435233760548].png results@4,[2.600974180557, 10.435233760548].png results@5,[2.600974180557, 10.435233760548].png results@6,[2.600974180557, 10.435233760548].png results@7,[2.600974180557, 10.435233760548].png results@8,[2.600974180557, 10.435233760548].png resultsK@0,[2.600974180557, 10.435233760548].png resultsK@1,[2.600974180557, 10.435233760548].png resultsK@2,[2.600974180557, 10.435233760548].png resultsK@3,[2.600974180557, 10.435233760548].png resultsK@4,[2.600974180557, 10.435233760548].png resultsK@5,[2.600974180557, 10.435233760548].png resultsK@6,[2.600974180557, 10.435233760548].png resultsK@7,[2.600974180557, 10.435233760548].png resultsK@8,[2.600974180557, 10.435233760548].png

README.md

@@ -1,50 +1,28 @@
-# 1. Challenge Overview:
-Vanguard’s portfolio construction process lies at the heart of its investment strategy, balancing risk, return, and investor preferences across a vast landscape of asset classes and constraints. However, as portfolios grow in complexity—spanning thousands of securities, intricate guardrails, and real-time trading demands—classical optimization tools like GUROBI face growing limitations in speed, scalability, and solution diversity. This challenge explores how sampling-based quantum optimization can be harnessed to overcome these barriers. By leveraging hybrid quantum-classical algorithms and decomposition pipelines, the goal is to prototype a quantum-enhanced solution that can:
+# Portfolio Optimization
 
-1. 	Efficiently solve high-dimensional, constraint-heavy portfolio optimization problems.
-2. 	Deliver near-optimal asset allocations within tight runtime windows.
-3. 	Scale to real-world use cases like fixed income ETF creation and index tracking.
-4. Preserve critical business metrics such as tracking error, excess return, and risk exposure.
+Project name: 
+* Annealed Portfolio
 
-The project focuses on using binary decision variables and quadratic objectives to simulate realistic trading scenarios. The challenge lies not only in achieving computational gains but also in maintaining interpretability, robustness, and alignment with investment principles. 
+Team Members:
+* Hao Mack Yang (gst-Z2od5sCatjVtRvH)
 
+Presentation:
+* [https://youtube.com/live/ZSvtRcaG_Vc](https://youtube.com/live/ZSvtRcaG_Vc)
 
-# 2. Challenge Duration:
- * 4 weeks
- * Teams start working on July 15, 2025
- * Teams submit their challenge solutions on August 10, 2025
+This **Washington Institute for STEM Entrepreneurship and Research** project demonstrates an alternative quantum computing method to optimize the portfolios based on the investor's preferences.
 
-# 3. Team Guidelines:
-*	Team size - Maximum 3 participants per team.
-*	All team participants must be enrolled in Womanium WISER Quantum 2025.
-*	Everyone is eligible to participate in this challenge and win Womanium grants.
-*	Best participants get selected for Womanium QSL fellowships with Vanguard.
+The goal is to minimize an objective function, the sum of the squared difference between the actual amount against the target amount of a target characteristic by risk bucket membership and characteristic while subject to guardrail constraints regarding the actual amounts. The minimization of this objective function allows us to maximize the expected utility of our portfolio. Our output is a set of binary decision variables indicating whether the particular bond is included in the portfolio.
 
-# 4. Challenge Tasks/ Deliverables:
-The participants are expected to complete for eligible challenge submission:
+Our approach uses the D-WAVE quantum annealing method, as opposed to the variational quantum algorithm implemented in Qiskit or Xanadu. Quantum annealing is especially well-suited for scalability and handling quadratic optimization problems, even with constraints. Unfortunately, we are not able to get access to the actual quantum processor through LEAP, so we ended up using one of D-WAVE's simulators. Additionally, constrained binary quadratic optimization problem still requires a hybrid approach.
 
-1)	Review the mathematical formulation provided below, focusing on binary decision variables, linear constraints, and the quadratic objective.
-2)	(necessary to pass the project) Convert the binary optimization problem to a formulation that is compatible with a quantum optimization algorithm. For example, convert the constrained problem to an unconstrained problem.
-3)	(necessary to pass the project) Write a quantum optimization program for handling problems of the type in (2). An example of such an optimization routine which is used in portfolio optimization is the Variational Quantum Eigensolver (see resources below), however you may pursue what you judge to be the best solution.
-4)	(challenge) Solve the optimization problem in (1) using your quantum formulation.
-5)	(challenge subtask) Validate your solution in (4) using a classical optimization routine. Compare the solution quality against the benchmark classical solution in terms of the cost function, and include relevant performance metrics(e.g., convergence of the optimization routine, and scaling properties with problem size).
-
-Note: No formal presentation is required. Instead, we’ll host a “show-and-tell” style session where each team will walk through their approach and demonstrate their prototype live. This is your opportunity to showcase your thinking, creativity, and results in an informal, interactive format.
+The structure of this submission is such that with elementary knowledge of Python or Jupyter plus the introduction of virtual environments, one can interactively see the process of data gathering, guardrail inputting, problem conversion, and simulated annealing, and presentation with one single click of the play button.
 
-# 5. Quantum Hardware Credits / Platform:
-*	Participants may use any quantum SDK or platform of their choice.
+With the exception of the README and presentations, all content that belongs to the team is in the `implementation` folder relative to the root working directory. Inside contains another README.md regarding the mathematical formulation, an `.ipynb` file depicting a large dataset run with randomly set bounds, a slightly modified `.ipynb` file from the `lp` bounds, a `.py` file depicting repeated runs of the small dataset runs, and the `bond_data1.json` generated for listing the selected attributes of the reduced dataset.
 
-# 6. Judging Criteria:
-Solutions will be evaluated against internal benchmark implementations at Vanguard. Evaluation will be based on:
-*	Speed of the solution
-*	Optimality (as measured by the cost function)
-*	Scalability (problem size handled)
+Upon running the script, a `dwave_result.csv` file will be generated at the `implementation` working directory (caveat: it is relative to the working directory of VS Code if you are running on VS Code), and various results files depicting the energy value, bond selection, and constraint satisfaction of each entry ordered by the D-WAVE energy rankings (the lowest energy selection is considered the best solution.)
 
-# 7. Resources:
+We have supplied the `result` and `result2` images, depicting the energy and secondary constraint satisfaction trends ranked by D-WAVE energy rating, and the `qaoamatrix.png` images, depicting the symmetrical matrix representing the binary QAOA matrix.
 
-*	An example of how VQE can be implemented - https://eric08000800.medium.com/portfolio-optimization-with-variational-quantum-eigensolver-vqe-2-477a0ee4e988
-*	A useful paper on variational Quantum Optimization https://quantum-journal.org/papers/q-2020-04-20-256/?utm_source=researcher_app&utm_medium=referral&utm_campaign=RESR_MRKT_Researcher_inbound
-*	Video recording for project orientation 2025 QUANTUM PROGRAM ❯ Day 7 ❯ Projects Orientation Part 2 - YouTube
-
-
+Please bear in mind that there are some limitations present in this project, especially concerning the potential differences between the actual boundaries used by the reference method against our method. The structure of the code is designed such that there should be minimal debugging and code redundancy when correcting the code to match the reference method to evaluate against.
 
+Word count: 460