Summary of Anole: Adapting Diverse Compressed Models For Cross-scene Prediction on Mobile Devices, by Yunzhe Li et al.
Anole: Adapting Diverse Compressed Models For Cross-Scene Prediction On Mobile Devices
by Yunzhe Li, Hongzi Zhu, Zhuohong Deng, Yunlong Cheng, Liang Zhang, Shan Chang, Minyi Guo
First submitted to arxiv on: 9 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
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 paper proposes Anole, a lightweight scheme for local deep neural network (DNN) model inference on mobile devices, particularly suitable for emerging AIoT applications. The key challenge addressed is the constant appearance of unfamiliar test samples due to device movement and unstable network connections. To overcome this, Anole establishes an army of compact DNN models and adaptively selects the best-fitting model for online inference. This is achieved through a weakly-supervised scene representation learning algorithm combining human heuristics and feature similarity, as well as a model classifier predicting the best-fit scene-specific DNN model. The results demonstrate Anole’s superiority over using a versatile large DNN in terms of prediction accuracy (4.5% higher), response time (33.1% faster) and power consumption (45.1% lower). |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Anole is a new way to make deep neural networks work better on mobile devices. These devices move around, so they often see things they haven’t seen before. This makes it hard for the network to make good predictions. Anole solves this problem by creating lots of small neural networks that can each predict something specific. It then chooses the best one for the job. The paper shows how Anole works and why it’s better than using a big, all-purpose neural network. |
Keywords
» Artificial intelligence » Inference » Neural network » Representation learning » Supervised
