Training Up to 50 Class ML Models on 3 $ IoT Hardware via Optimizing One-vs-One Algorithm (Student Abstract)

Multi-class classifier training using traditional meta-algorithms such as the popular One-vs-One (OvO) method may not always work well under cost-sensitive setups. Also, during inference, OvO becomes computationally challenging for higher class counts K O(K^2) is its time complexity. In this paper, we present Opt-OvO, an optimized (resource-friendly) version of algorithm to enable high-performance multi-class ML and inference directly on microcontroller units (MCUs). Opt-OvO enables billions tiny IoT devices self learn/train (offline) after their deployment, live data from a wide range use-cases. We demonstrate by performing model 4 MCU boards datasets varying counts, sizes, feature dimensions. The most exciting finding was, 3 $ ESP32 chip, trained dataset count 50 performed unit in super real-time 6.2 ms.