Shiqing Gao

• Completed a reproducible undergraduate computational study of denoising and deblurring under controlled synthetic degradations.
• Compared Gaussian filtering, Tikhonov regularization, TV denoising, non-local means, Wiener deconvolution, Richardson-Lucy deconvolution, and a small CNN baseline.
• Evaluated PSNR, SSIM, runtime, visual comparisons, error maps, parameter sensitivity, and degradation robustness.

Tools: Python, NumPy, pandas, matplotlib, scikit-image

• Completed a toy spectral visualization comparing Fourier mode decay in normal diffusion and time-fractional subdiffusion.
• Used a sine eigenbasis on Omega = (0, 1) with exponential and Mittag-Leffler-type decay factors.
• Documented the limitation that the project is a visualization, not a full inverse solver.

Tools: Python, NumPy, matplotlib

• Completed a reproducible regression uncertainty project using the UCI Energy Efficiency dataset with Heating Load as the target.
• Compared approximate Gaussian process intervals, split conformal prediction, naive bootstrap intervals, and residual-corrected bootstrap intervals.
• Studied empirical coverage, interval width, repeated-split stability, reduced-training-size behavior, and perturbed-input robustness.

Tools: Python, NumPy, pandas, scikit-learn, matplotlib

• Completed a chip thermal management modeling workflow using surrogate models for thermal resistance, pressure drop, and temperature nonuniformity.
• Used Pareto filtering, weighted-sum ranking, TOPSIS, weight-scenario analysis, random weight sampling, and Monte Carlo perturbation checks.
• Produced tables, figures, and written analysis focused on assumptions, trade-offs, robustness, and limitations.

Tools: Python, NumPy, pandas, matplotlib, LaTeX