Timezone: »

 
Poster
Predicting Training Time Without Training
Luca Zancato · Alessandro Achille · Avinash Ravichandran · Rahul Bhotika · Stefano Soatto

Thu Dec 10 09:00 AM -- 11:00 AM (PST) @ Poster Session 5 #1655
in Poster Session 6 »

We tackle the problem of predicting the number of optimization steps that a pre-trained deep network needs to converge to a given value of the loss function. To do so, we leverage the fact that the training dynamics of a deep network during fine-tuning are well approximated by those of a linearized model. This allows us to approximate the training loss and accuracy at any point during training by solving a low-dimensional Stochastic Differential Equation (SDE) in function space. Using this result, we are able to predict the time it takes for Stochastic Gradient Descent (SGD) to fine-tune a model to a given loss without having to perform any training.
In our experiments, we are able to predict training time of a ResNet within a 20\% error margin on a variety of datasets and hyper-parameters, at a 30 to 45-fold reduction in cost compared to actual training. We also discuss how to further reduce the computational and memory cost of our method, and in particular we show that by exploiting the spectral properties of the gradients' matrix it is possible to predict training time on a large dataset while processing only a subset of the samples.

Author Information

Luca Zancato (University of Padova)
Alessandro Achille (Amazon Web Services)
Avinash Ravichandran (AWS)
Rahul Bhotika (Amazon Web Services)
Stefano Soatto (UCLA)

More from the Same Authors