Complex Deep Neural Networks such as Capsule Networks (CapsNets) exhibit high learning capabilities at the cost of compute-intensive operations. To enable their deployment on edge devices, we propose to leverage approximate computing for designing approximate variants of the complex operations like softmax and squash. In our experiments, we evaluate tradeoffs between area, power consumption, and critical path delay of the designs implemented with the ASIC design flow, and the accuracy of the quantized CapsNets, compared to the exact functions.
Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations / Marchisio, Alberto; Bussolino, Beatrice; Salvati, Edoardo; Martina, Maurizio; Masera, Guido; Shafique, Muhammad. - ELETTRONICO. - (2022). (Intervento presentato al convegno ISLPED '22: ACM/IEEE International Symposium on Low Power Electronics and Design tenutosi a Boston (USA) nel 1-2 August 2022) [10.1145/3531437.3539717].
Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations
Bussolino, Beatrice;Martina, Maurizio;Masera, Guido;
2022
Abstract
Complex Deep Neural Networks such as Capsule Networks (CapsNets) exhibit high learning capabilities at the cost of compute-intensive operations. To enable their deployment on edge devices, we propose to leverage approximate computing for designing approximate variants of the complex operations like softmax and squash. In our experiments, we evaluate tradeoffs between area, power consumption, and critical path delay of the designs implemented with the ASIC design flow, and the accuracy of the quantized CapsNets, compared to the exact functions.File | Dimensione | Formato | |
---|---|---|---|
ispled2022_editoriale.pdf
accesso aperto
Descrizione: Versione editoriale
Tipologia:
2a Post-print versione editoriale / Version of Record
Licenza:
Pubblico - Tutti i diritti riservati
Dimensione
1.55 MB
Formato
Adobe PDF
|
1.55 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/2968137