Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2509.18121 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918262575464448 |
|---|---|
| author | Garg, Nikhil Vicandi, Paul Uriarte Zhang, Yanming Baigol, Alexandre Falcone, Donato Francesco Mamidala, Saketh Ram Offrein, Bert Jan Bégon-Lours, Laura |
| author_facet | Garg, Nikhil Vicandi, Paul Uriarte Zhang, Yanming Baigol, Alexandre Falcone, Donato Francesco Mamidala, Saketh Ram Offrein, Bert Jan Bégon-Lours, Laura |
| contents | The increasing deployment of wearable sensors and implantable devices is shifting AI processing demands to the extreme edge, necessitating ultra-low power for continuous operation. Inspired by the brain, emerging memristive devices promise to accelerate neural network training by eliminating costly data transfers between compute and memory. Though, balancing performance and energy efficiency remains a challenge. We investigate ferroelectric synaptic devices based on HfO2/ZrO2 superlattices and feed their experimentally measured weight updates into hardware-aware neural network simulations. Across pulse widths from 20 ns to 0.2 ms, shorter pulses lower per-update energy but require more training epochs while still reducing total energy without sacrificing accuracy. Classification accuracy using plain stochastic gradient descent (SGD) is diminished compared to mixed-precision SGD. We analyze the causes and propose a ``symmetry point shifting'' technique, addressing asymmetric updates and restoring accuracy. These results highlight a trade-off among accuracy, convergence speed, and energy use, showing that short-pulse programming with tailored training significantly enhances on-chip learning efficiency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_18121 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Energy-convergence trade off for the training of neural networks on bio-inspired hardware Garg, Nikhil Vicandi, Paul Uriarte Zhang, Yanming Baigol, Alexandre Falcone, Donato Francesco Mamidala, Saketh Ram Offrein, Bert Jan Bégon-Lours, Laura Emerging Technologies Machine Learning Systems and Control The increasing deployment of wearable sensors and implantable devices is shifting AI processing demands to the extreme edge, necessitating ultra-low power for continuous operation. Inspired by the brain, emerging memristive devices promise to accelerate neural network training by eliminating costly data transfers between compute and memory. Though, balancing performance and energy efficiency remains a challenge. We investigate ferroelectric synaptic devices based on HfO2/ZrO2 superlattices and feed their experimentally measured weight updates into hardware-aware neural network simulations. Across pulse widths from 20 ns to 0.2 ms, shorter pulses lower per-update energy but require more training epochs while still reducing total energy without sacrificing accuracy. Classification accuracy using plain stochastic gradient descent (SGD) is diminished compared to mixed-precision SGD. We analyze the causes and propose a ``symmetry point shifting'' technique, addressing asymmetric updates and restoring accuracy. These results highlight a trade-off among accuracy, convergence speed, and energy use, showing that short-pulse programming with tailored training significantly enhances on-chip learning efficiency. |
| title | Energy-convergence trade off for the training of neural networks on bio-inspired hardware |
| topic | Emerging Technologies Machine Learning Systems and Control |
| url | https://arxiv.org/abs/2509.18121 |