Fractional Max-Pooling

Training compute

10¹⁷ FLOP

Parameters

27M

Published

Dec 18, 2014

Fractional Max-Pooling is an AI model developed by University of Warwick (United Kingdom), first published in December 2014. It works in the vision domain, on tasks such as image classification.

Training it took an estimated 10¹⁷ FLOP of compute (estimation method: hardware). The model has 27,000,000 parameters. It was trained on roughly 901.2K datapoints. Training ran on NVIDIA GeForce GTX 780 for about 18 hours.

The reference paper has 672 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: University of Warwick
Country of organization: United Kingdom
Domain: Vision
Task: Image classification
Training compute: 10¹⁷ FLOP
Compute estimation method: Hardware
Parameters: 27,000,000
Dataset size: 901.2K
Training hardware: NVIDIA GeForce GTX 780
Training time: 18 h
Citations: 672
Epoch confidence: Likely