Deep Autoencoders

Training compute

3.7×10¹⁶ FLOP

Parameters

139.8M

Published

Apr 29, 2011

Deep Autoencoders is an AI model developed by University of Toronto (Canada), first published in April 2011. It works in the vision domain, on tasks such as image representation. It counts among the frontier models: the systems trained with the most compute of their moment.

Training it took an estimated 3.7×10¹⁶ FLOP of compute (estimation method: hardware). The model has 139,808,256 parameters. It was trained on roughly 4.9B datapoints. Training ran on 1 NVIDIA GeForce GTX 285 for about 48 hours.

Epoch AI rates the confidence of this record as confident.

Full record

Organization: University of Toronto
Country of organization: Canada
Domain: Vision
Task: Image representation
Training compute: 3.7×10¹⁶ FLOP
Compute estimation method: Hardware
Parameters: 139,808,256
Dataset size: 4.9B
Training hardware: NVIDIA GeForce GTX 285
Chips used: 1
Training time: 48 h
Training power draw: 246 W
Epoch confidence: Confident