We present AnaCoNGA, an analytical co-design methodology, which enables two genetic algorithms to evaluate the fitness of design decisions on layer-wise quantization of a neural network and hardware (HW) resource allocation. We embed a hardware architecture search (HAS) algorithm into a quantization strategy search (QSS) algorithm to evaluate the hardware design Pareto-front of each considered quantization strategy. We harness the speed and flexibility of analytical HW-modeling to enable parallel HW-CNN co-design. With this approach, the QSS is focused on seeking high-accuracy quantization strategies which are guaranteed to have efficient hardware designs at the end of the search. Through AnaCoNGA, we improve the accuracy by 2.88 p.p. with respect to a uniform 2-bit ResNet20 on CIFAR-10, and achieve a 35% and 37% improvement in latency and DRAM accesses, while reducing LUT and BRAM resources by 9% and 59% respectively, when compared to a standard edge variant of the accelerator. The nested genetic algorithm formulation also reduces the search time by 51% compared to an equivalent, sequential co-design formulation.

AnaCoNGA: Analytical HW-CNN Co-Design Using Nested Genetic Algorithms / Fasfous, Nael; Vemparala, Manoj Rohit; Frickenstein, Alexander; Valpreda, Emanuele; Salihu, Driton; Hufer, Julian; Singh, Anmol; Nagaraja, Naveen-Shankar; Voegel, Hans-Joerg; Vu Doan, Nguyen Anh; Martina, Maurizio; Becker, Juergen; Stechele, Walter. - ELETTRONICO. - (2022), pp. 238-243. ((Intervento presentato al convegno 2022 Design, Automation & Test in Europe Conference & Exhibition (DATE) tenutosi a Antwerp, Belgium nel 14-23 March 2022 [10.23919/DATE54114.2022.9774574].

AnaCoNGA: Analytical HW-CNN Co-Design Using Nested Genetic Algorithms

Valpreda, Emanuele;Martina, Maurizio;
2022

Abstract

We present AnaCoNGA, an analytical co-design methodology, which enables two genetic algorithms to evaluate the fitness of design decisions on layer-wise quantization of a neural network and hardware (HW) resource allocation. We embed a hardware architecture search (HAS) algorithm into a quantization strategy search (QSS) algorithm to evaluate the hardware design Pareto-front of each considered quantization strategy. We harness the speed and flexibility of analytical HW-modeling to enable parallel HW-CNN co-design. With this approach, the QSS is focused on seeking high-accuracy quantization strategies which are guaranteed to have efficient hardware designs at the end of the search. Through AnaCoNGA, we improve the accuracy by 2.88 p.p. with respect to a uniform 2-bit ResNet20 on CIFAR-10, and achieve a 35% and 37% improvement in latency and DRAM accesses, while reducing LUT and BRAM resources by 9% and 59% respectively, when compared to a standard edge variant of the accelerator. The nested genetic algorithm formulation also reduces the search time by 51% compared to an equivalent, sequential co-design formulation.
File in questo prodotto:
File Dimensione Formato  
2022_DATE___AnaCoNGA_Accepted_.pdf

accesso aperto

Descrizione: Articolo accettato prima di modifiche editore
Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 610.86 kB
Formato Adobe PDF
610.86 kB Adobe PDF Visualizza/Apri
AnaCoNGA_Analytical_HW-CNN_Co-Design_Using_Nested_Genetic_Algorithms.pdf

non disponibili

Descrizione: Versione finale pubblicata sui proceedings
Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 697.04 kB
Formato Adobe PDF
697.04 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11583/2964390