Summary of Ikan: Global Incremental Learning with Kan For Human Activity Recognition Across Heterogeneous Datasets, by Mengxi Liu et al.
iKAN: Global Incremental Learning with KAN for Human Activity Recognition Across Heterogeneous Datasets
by Mengxi Liu, Sizhen Bian, Bo Zhou, Paul Lukowicz
First submitted to arxiv on: 3 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
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 The proposed incremental learning (IL) framework, iKAN, tackles catastrophic forgetting and non-uniform inputs for wearable sensor human activity recognition (HAR). iKAN pioneers IL with Kolmogorov-Arnold Networks (KAN) as the classifier, leveraging local plasticity and global stability of splines. The scalable framework adapts KAN for HAR using task-specific feature branches and a feature redistribution layer. Unlike existing IL methods, iKAN expands feature extraction branches to accommodate new inputs from different sensor modalities while maintaining consistent dimensions and output numbers. The framework demonstrates incremental learning performance across six public HAR datasets with a last performance of 84.9% (weighted F1 score) and an average incremental performance of 81.34%, outperforming existing methods such as EWC (51.42%) and experience replay (59.92%). |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary iKAN is a new way to learn from data collected by wearable sensors, like smartwatches or fitness trackers. It helps machines remember what they learned before, even when the type of data changes. This is important because most machines forget old information when learning new things. The iKAN method uses special kinds of neural networks and takes into account the different types of sensor data collected. It’s tested on six different datasets and works better than existing methods. |
Keywords
» Artificial intelligence » Activity recognition » F1 score » Feature extraction
