We investigate novel SoC-FPGA solutions for fast and energy-efficient ranking based on machine-learned ensembles of decision trees. Since the memory footprint of ranking ensembles limits the effective exploitation of programmable logic for large-scale inference tasks, we investigate binning and quantization techniques to reduce the memory occupation of the learned model and we optimize the state-of-the-art ensemble-traversal algorithm for deployment on low-cost, energy-efficient FPGA devices. The results of the experiments conducted using publicly available Learning-to-Rank datasets, show that our model compression techniques do not impact significantly the accuracy. Moreover, the reduced space requirements allow the models and the logic to be replicated on the FPGA device in order to execute several inference tasks in parallel. We discuss in details the experimental settings and the feasibility of the deployment of the proposed solution in a real setting. The results of the experiments conducted show that our FPGA solution achieves performances at the state of the art and consumes from 9x up to 19.8x less energy than an equivalent multi-threaded CPU implementation.

Ensemble Model Compression for~Fast and~Energy-Efficient Ranking on~{FPGAs}

Romina Molina;
2022-01-01

Abstract

We investigate novel SoC-FPGA solutions for fast and energy-efficient ranking based on machine-learned ensembles of decision trees. Since the memory footprint of ranking ensembles limits the effective exploitation of programmable logic for large-scale inference tasks, we investigate binning and quantization techniques to reduce the memory occupation of the learned model and we optimize the state-of-the-art ensemble-traversal algorithm for deployment on low-cost, energy-efficient FPGA devices. The results of the experiments conducted using publicly available Learning-to-Rank datasets, show that our model compression techniques do not impact significantly the accuracy. Moreover, the reduced space requirements allow the models and the logic to be replicated on the FPGA device in order to execute several inference tasks in parallel. We discuss in details the experimental settings and the feasibility of the deployment of the proposed solution in a real setting. The results of the experiments conducted show that our FPGA solution achieves performances at the state of the art and consumes from 9x up to 19.8x less energy than an equivalent multi-threaded CPU implementation.
File in questo prodotto:
File Dimensione Formato  
Matthias Hagen (editor), Suzan Verberne (editor), Craig Macdonald (editor), Christin Seifert (editor), Krisztian Balog (editor), Kjetil Nørvåg (editor), Vinay Setty (editor) - Advances in Information -1.pdf

Accesso chiuso

Descrizione: cover. index. chapter
Tipologia: Documento in Versione Editoriale
Licenza: Copyright Editore
Dimensione 481.82 kB
Formato Adobe PDF
481.82 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
Matthias+Hagen+(editor),+Suzan+Verberne+(editor),+Craig+Macdonald+(editor),+Christin+Seifert+(editor),+Krisztian+Balog+(editor),+Kjetil+Nørvåg+(editor),+Vinay+.pdf

Open Access dal 06/04/2023

Tipologia: Bozza finale post-referaggio (post-print)
Licenza: Digital Rights Management non definito
Dimensione 899.48 kB
Formato Adobe PDF
899.48 kB Adobe PDF Visualizza/Apri
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: https://hdl.handle.net/11368/3034438
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 4
  • ???jsp.display-item.citation.isi??? 2
social impact