Putting intelligence on chips small enough to be in disconcerting places

October 14, 2016 — August 14, 2023

bounded compute
edge computing
machine learning
neural nets
sparser than thou
Figure 1

The art of doing ML stuff on small controllers, such as single board computers or even individual microcontrollers, or some weird chip embedded in some medical device. A.k.a. edge ml,1

I do not have much to say here RN.

1 Making neural models small

Obviously if your target model is a neural net one important step is making it be as small as possible in the sense of having as few parameters as possible.

2 Low precision nets

There is another sense of small: Using 16 bit float, fixed point, or even single bit, arithmetic so that the numbers involved are compact. TBD.

3 Multi-task learning

Training one learning algorithm to solve several problems simultaneously. Probably needs its own page.

4 Tooling

Tensorflow appears to have intimate microcontroller integration via TensorFlow Lite for Microcontrollers.

Browser ML in particular has some quirks.

Other frameworks convert to intermediate format ONNX which can be run on microcontrollers, although I suspect with higher overhead.

5 Minifying neural nets

6 compiling neural nets

8 References

Cai, Gan, Wang, et al. 2020. Once-for-All: Train One Network and Specialize It for Efficient Deployment.” In.
Chen, Goodfellow, and Shlens. 2015. Net2Net: Accelerating Learning via Knowledge Transfer.” arXiv:1511.05641 [Cs].
Cheng, Wang, Zhou, et al. 2017. A Survey of Model Compression and Acceleration for Deep Neural Networks.” arXiv:1710.09282 [Cs].
Cohn, Agarwal, Gupta, et al. 2023. EELBERT: Tiny Models Through Dynamic Embeddings.” In.
He, Lin, Liu, et al. 2019. AMC: AutoML for Model Compression and Acceleration on Mobile Devices.” arXiv:1802.03494 [Cs].
Howard, Zhu, Chen, et al. 2017. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications.” arXiv:1704.04861 [Cs].
Roth. 2021. Probabilistic Methods for Resource Efficiency in Machine Learning.”
Shi, Feng, and ZhifanZhu. 2016. Functional Hashing for Compressing Neural Networks.” arXiv:1605.06560 [Cs].
Waltsburger, Libessart, Ren, et al. 2023. Neural Network Scoring for Efficient Computing.” In 2023 IEEE International Symposium on Circuits and Systems (ISCAS).
Wang, Xu, Xu, et al. 2019. Packing Convolutional Neural Networks in the Frequency Domain.” IEEE transactions on pattern analysis and machine intelligence.
Warden, and Situnayake. 2020. TinyML: Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-Power Microcontrollers.


  1. I do not like this term, because it tends to imply that we care especially about some kind of centre-edge distinction, which we only do sometimes. It tends to imply that large NN models in data centres are the default type of ML. Chris Mountford’s Hasn’t AI Been the Wrong Edgy for Too Long?, mentioned in the comments riffs on this harder than I imagined, though↩︎