Machine Learning Engineering
Null-Space Projected PIELM for Exact Constraint Enforcement
Physics-informed extreme learning machines (PIELMs) traditionally use penalty terms for boundary and initial conditions, leading to approximate satisfaction and error propagation. This paper introduces Null-Space Projected PIELM (NP-PIELM), which employs an algebraic projection in coefficient space …
INT4 Quantization Enables Efficient Federated Learning for Aerospace Predictive Maintenance
Federated learning for aerospace predictive maintenance is challenged by bandwidth limitations. This paper demonstrates that INT4 quantization, applied to a lightweight 1-D convolutional model (AeroConv1D), achieves accuracy comparable to FP32 while significantly reducing communication overhead. The…
Operationalizing ML: Transitioning from Model Training to Production Lifecycle Management
The updated Machine Learning in Production course shifts focus from isolated model training to the holistic deployment lifecycle. It provides a framework for project scoping, data management, and operational maintenance to ensure robust model performance in real-world applications.