hon-gyu,Hongyu Lin,github

Hi there 👋

🔍 I'm Hongyu, a passionate quant particularly interested in emerging machine learning applications.

🌍 Based in London. Open to work.

🗣️ Fluent in Mandarin & English.

Skills

Programming Language: Python, R, SQL
Machine Learning
- Frameworks: PyTorch, Scikit-Learn, Keras
- Deep Learning: LSTM
- Model Interpretation: Shap, LIME
- Clustering: Hierarchical Clustering, Chameleon Clustering, Semi-supervised Clustering
- Ensemble Trees: Gradient-Boosted Trees, Random Forest
- Conformal Prediction: MAPIE
- Reinforcement Learning
Others: Git, Linux, LaTex

Projects

Semi-supervised Chameleon Clustering: An implementation of semi-supervised Chameleon clustering, capable of integrating must-link and cannot-link constraints at various levels of hierarchy to guide the clustering process.
Model Fingerprint: A model-agnostic method to decompose predictions into linear, nonlinear and pairwise interaction effects. It can be helpful in feature selection and model interpretation.
Using LSTM Model for Meta-labeling: An implementation that applies meta-labeling to minute-frequency stock data, utilizing LSTM as the primary model for price direction prediction, which forms the basis for a trading strategy augmented by a secondary meta-labeling layer to filter false positives and improve risk-return metrics.
Semi-supervised Hedge Fund Clustering: A semi-supervised clustering method utilizing tree distance for enhanced hierarchical classification of funds in fund of funds analysis.
Hierarchical Tree Distance: An implementation of the AKB tree distance, a measure designed to quantify the similarity between classes within a hierarchical label tree. Adept at emphasizing the importance of higher hierarchy errors, utilizing the taxonomy's inherent structure instead of simply flattening the hierarchy in traditional.

Micro Projects

PyTorch Model Interpretation by Shap: An implemention of using PyTorch model in shap framework, which is a game theoretic approach to explain the output of any machine learning model. Created shap.Explanation object for PyTorch models to facilitate visualisation using a unified interface.

Hongyu Lin's Projects

chameleon_cluster

A chameleon clustering algorithm implementation in PYTHON3.

DoWhy is a Python library for causal inference that supports explicit modeling and testing of causal assumptions. DoWhy is based on a unified language for causal inference, combining causal graphical models and potential outcomes frameworks.

feature-clustering

A data-driven dimensionality reduction analysis utilizing Dask and Scikit-Learn to be used in a paper for Lawrence Berkeley National Laboratory

hierarchical-tree-distance

An implementation of the AKB tree distance, a measure designed to quantify the similarity between classes within a hierarchical label tree. Adept at emphasizing the importance of higher hierarchy errors, utilizing the taxonomy's inherent structure instead of simply flattening the hierarchy in traditional.

langchain

⚡ Building applications with LLMs through composability ⚡

meta-labeling-and-lstm

An implementation that applies meta-labeling to minute-frequency stock data, utilizing LSTM as the primary model for price direction prediction, which forms the basis for a trading strategy augmented by a secondary meta-labeling layer to filter false positives and improve risk-return metrics.

model-fingerprint

A model-agnostic method to decompose predictions of machine learning models into linear, nonlinear and pairwise interaction effects.

pytorch-shap

An implemention of using PyTorch model in shap framework. Created shap.Explanation object for PyTorch models to facilitate visualisation using a unified interface.

risklabai.py

Financial AI with Python

semi-supervised-chameleon-clustering

An implementation of semi-supervised Chameleon clustering, capable of integrating must-link and cannot-link constraints at various levels of hierarchy to guide the clustering process. It also offers the ability to perform Chameleon clustering without any constraints, i.e., pure unsupervised clustering.