Machine Learning (Python)

Yunhui Hu

***** ******** ******, **, ***** 949-***-**** ******@****.***

EDUCATION

University of California San Diego, Irwin & Joan Jacobs School of Engineering ECE - Machine learning and Control depth Sep 2023 - now SKILLS

Programming languages: C, C++, Python, R language (Basic), Matlab Machine learning tools: Scikit-learn, Numpy, Pandas, Matplotlib, Seaborn Other technical skills: Git, Jupyter Notebook

Mathematical & Analytical knowledge: Probability Theory (Engineering Applications), Mathematical Statistics, Linear & Logistic Regression, Support Vector Machine, Statistical Modeling & Inference, Hypothesis Testing, K-NN, Decision Tree, Classification Using Generative Models, Linear and Nonlinear Optimization with Applications

Project

Multi-Class Classification on MNIST using Least Squares

● Built one-vs-all and one-vs-one classifier from scratch in Python to recognize handwritten digits from the MNIST dataset using least squares optimization

● Preprocessed and normalized 60,000 training and 10,000 test images; evaluated classifiers using error rate and confusion matrices.

● Extended model with randomized nonlinear feature maps (Sigmoid, ReLU, Sinusoidal, Identity) to capture non-linear patterns; significantly improved test accuracy.

● Assessed classifier robustness under bounded noise corruption, identifying performance degradation thresholds for each feature mapping.

Neural Network Training with Nonlinear Least Squares

● Implemented a fully connected neural network with custom backpropagation using Python, trained using the Levenberg–Marquardt algorithm to approximate nonlinear functions.

● Derived and coded custom gradient expressions using tanh activation, and minimized training loss via iterative optimization.

● Conducted experiments on regularization (λ tuning), initialization sensitivity, and impact of noise levels in the data on model performance.

● Evaluated generalization on unseen test samples and visualized learned surfaces with contour plots to interpret network accuracy.

Exploratory Data Science Project on sleeping quality

● Identified a socially and scientifically relevant question, collected and cleaned data, and performed exploratory data analysis (EDA) to uncover trends and insights.

● Developed and tested models using techniques such as feature engineering, data visualization.

● Addressed ethical and privacy considerations in data sourcing and interpretation.

Contact this candidate