A Survey with Practical Demonstration on Acceleration of Machine Learning Models

The rapid advancement of machine learning (ML) has led to an increasing demand for efficient computation methods to handle ML models in real-time applications. This paper reviews recent methods for acceleration of machine learning models. We focus on techniques that enhance performance while optimizing resource utilization. We explore various hardware platforms, including GPUs, FPGAs, and ASICs, and discuss their respective advantages and limitations in accelerating machine learning workloads. Furthermore, we demonstrate a framework for using FPGAs for paper scissor rock classification. Our findings suggest that while significant progress has been made, more innovative methods are necessary to facilitate hardware capabilities for ML models.