Optimizing Neural Networks for TinyML

  Machine learning


Embedded devices have dedicated acceleration hardware for machine learning such as parallel instruction sets for CPUs, multi-purpose GPUs with vision kernels, Digital Signal Processors for audio processing, or Neural Processing Units specialized to run neural networks. All of this hardware requires neural network models that are optimized using techniques such as weight clustering, pruning, and most importantly quantization. We will discuss details of such optimizations for different use cases, different hardware, their impact on performance, and the accuracy trade-off.

17:30 Welcome chat
18:00 Talk
18:45 Discussion
19:00 Networking (Impact Hub)

Machine Learning Meetups (MLMU) is an independent platform for people interested in Machine Learning, Information Retrieval, Natural Language Processing, Computer Vision, Pattern Recognition, Data Journalism, Artificial Intelligence, Agent Systems and all the related topics. MLMU is a regular community meeting usually consisting of a talk, a discussion and subsequent networking. Except of Prague, MLMU also spread to Brno, Bratislava and Košice.


Pavel Macenauer

Pavel Macenauer is an engineering lead at NXP Semiconductors for AI and Machine Learning where he enables inference engines and neural network-related tooling for a wide range of microprocessors and microcontrollers. Throughout his carreer as a software engineer he developed applications and libraries ranging from graphics and photography, through computer vision to machine learning for embedded and safety-critical systems. Apart from that he spends his spare time with photography, lectures image post-processing, and co-founded FotoInstitut.cz, FotoAparat.cz, and a few other photo-related projects.