InTreeger: An End-to-End Framework for Integer-Only Decision Tree Inference

2025-05-21

AI Summary

To address accuracy degradation and energy-efficiency bottlenecks in decision-tree inference on resource-constrained devices, this paper proposes the first end-to-end integer-quantized decision-tree compilation framework. The method generates architecture-agnostic, zero-accuracy-loss C code comprising purely integer arithmetic—requiring no floating-point unit support and imposing no structural constraints on input trees. Its core contributions include: (1) a provably lossless integer quantization compilation algorithm; (2) lightweight, tree-structure-aware optimizations; and (3) automatic low-level C code synthesis targeting ARM, x86, and RISC-V instruction sets. Experimental evaluation across diverse edge platforms demonstrates an average 3.2× reduction in inference latency and up to 5.7× improvement in energy efficiency. The framework successfully deploys on sub-milliwatt embedded systems, validating its suitability for ultra-low-power applications.

Technology Category

• Machine Learning: Hardware-aware ML
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Computer Vision: Learning & Optimization for CV

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Integer quantization has emerged as a critical technique to facilitate deployment on resource-constrained devices. Although they do reduce the complexity of the learning models, their inference performance is often prone to quantization-induced errors. To this end, we introduce InTreeger: an end-to-end framework that takes a training dataset as input, and outputs an architecture-agnostic integer-only C implementation of tree-based machine learning model, without loss of precision. This framework enables anyone, even those without prior experience in machine learning, to generate a highly optimized integer-only classification model that can run on any hardware simply by providing an input dataset and target variable. We evaluated our generated implementations across three different architectures (ARM, x86, and RISC-V), resulting in significant improvements in inference latency. In addition, we show the energy efficiency compared to typical decision tree implementations that rely on floating-point arithmetic. The results underscore the advantages of integer-only inference, making it particularly suitable for energy- and area-constrained devices such as embedded systems and edge computing platforms, while also enabling the execution of decision trees on existing ultra-low power devices.

Problem

• Reduces quantization-induced errors in integer-only decision tree inference
• Generates architecture-agnostic integer-only C code without precision loss
• Improves energy efficiency for resource-constrained embedded and edge devices

Innovation

• End-to-end integer-only decision tree framework
• Generates architecture-agnostic C implementation
• Optimized for energy-constrained edge devices