Summary of 12 Mj Per Class On-device Online Few-shot Class-incremental Learning, by Yoga Esa Wibowo et al.
12 mJ per Class On-Device Online Few-Shot Class-Incremental Learning
by Yoga Esa Wibowo, Cristian Cioflan, Thorir Mar Ingolfsson, Michael Hersche, Leo Zhao, Abbas Rahimi, Luca Benini
First submitted to arxiv on: 12 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes Online Few-Shot Class-Incremental Learning (O-FSCIL), a lightweight machine learning approach that enables devices at the extreme edge to learn from only a few labeled examples without forgetting previously learned classes. The O-FSCIL architecture consists of a pretrained and metalearned feature extractor, an explicit memory storing class prototypes, and a novel feature orthogonality regularization. This allows for online learning of new classes with a single pass. The approach is tested on the FSCIL CIFAR100 benchmark, achieving state-of-the-art results with an average accuracy of 68.62%. Additionally, O-FSCIL is implemented on the low-power GAP9 microcontroller, demonstrating online learning capabilities within just 12 mJ per new class. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary O-FSCIL is a special type of machine learning that helps devices learn from very few examples without forgetting what they already know. This is important for devices like smartphones or cameras that need to be able to recognize things even when they don’t have much power. The O-FSCIL approach uses a simple and efficient model that can be trained using only a few labeled examples, which makes it perfect for use on low-power devices. |
Keywords
* Artificial intelligence * Few shot * Machine learning * Online learning * Regularization
