🌎 Data Scientist | ML Researcher | AI for Sustainability
- 🚀 Passionate about AI for Sustainability, Machine Learning, and Remote Sensing.
- 🌱 Exploring cutting-edge deep learning methods for real-world impact.
- 🔭 Researching adversarial robustness, AI for Earth observation, and computational optimization.
- ✨ Love working on computer vision, NLP, and AI for social good.
-
🛰 AI for Satellite-based Monitoring (Canadian Space Agency, European Space Agency)
- Developing robust AI models for remote sensing, crisis detection, and environmental monitoring.
- Enhancing adaptability in adverse conditions (e.g., nightlight super-resolution for crisis mapping).
-
🚗 Autonomous Driving Safety (ConvAE for Adverse Weather Detection)
- Optimizing ConvAE for robustness in adverse weather detection.
- Improving model generalization across diverse environmental conditions for safer self-driving systems.
-
🛰 Machine Learning for Space Applications (Canadian Space Agency, European Space Agency)
- Applied deep learning & computer vision to automate satellite-based data processing.
- Investigated AI-driven prediction models for environmental monitoring and disaster response.
-
🔬 Adversarially Robust Memory Systems: Restricted Hopfield Networks (RHN)
- The Restricted Hopfield Network (RHN) enhances Hopfield Neural Networks (HNN) by introducing hidden layers and leveraging the Subspace Rotation Algorithm (SRA) for better robustness and storage capacity.
- RHN outperforms Dense Associative Memory (DAM), Predictive Coding Network (PCN), and MLP in adversarial scenarios, showing resilience against FGSM, BIM, PGD, and Gaussian Noise attacks while ensuring high retrieval accuracy.
💡 Tech Stack:
- 🧠 Deep Learning: PyTorch, TensorFlow, Detectron2, MMDetection
- 📊 Data Science: Python, R, PostgreSQL, MATLAB, GIS
- 🚀 Cloud & Infra: Azure, Databricks, Kubernetes
📄 Enhancing Safety of Autonomous Vehicles in Adverse Weather
📄 AI-Driven Predictive Models for Sustainable Agriculture
📄 Restricted Hopfield Networks and Adversarial Robustness
📄 Boosting Adversarial Robustness with Gradient-Feature Alignment (Under Review at the 38th Canadian Conference on AI)
📄 Bidirectional Associative Memory Trained by Subspace Rotation Algorithm (Under Review at IJCAI 2025)
📄 Enhancing Adversarial Resilience in Ensembles Through Combined Gradient Alignment and Gradient Norm Regularization (Under Review at IJCAI 2025)
📄 Restricted Hopfield Networks are Robust to Adversarial Attack (Manuscript in preparation / Preprint forthcoming)
📩 Email: bzhan138@uottawa.ca
🌐 Website: yourwebsite.com 🚧 (Coming Soon...)
🐦 Twitter: @yourhandle 🚧 (Coming Soon...)
⭐ Fun Fact: I love urban aesthetics, museums, and writing papers on cutting-edge AI methods! 🤖📚
