- This event has passed.
Surprises in Deep Learning Training – Learning Club talk by Daniel Soudry, Technion
June 21 @ 12:00 pm - 1:00 pm IDT
On Sunday 21.6 12:00 we will host Daniel Soudry from the Technion in our machine learning seminar.
The talk will be about new findings in training deep neural networks, from a number of recent publications by Daniel’s group (see more details below).
Due to the coronavirus the talk will take place via zoom: https://zoom.us/j/
See you all soon,
Title: Surprises in Deep Learning Training
Abstract: We make a few interesting observations regarding deep neural networks (DNNS) training:
1) DNNs are typically initialized with (a) random weights in order to (b) break symmetry and (c) promote feature diversity.
We demonstrate a,b, and c are not necessary at initialization to obtain high accuracy at the end of the training. (ICML2020)
2) Large batch training is commonly used to accelerate training.
We improve final accuracy by increasing the batch size with more data augmentations, instead of more samples. (CVPR2020)
3) Quantization of full precision trained DNNs while retaining high accuracy typically requires fine-tuning the model on a large dataset.
We generate synthetic data from the DNN Batch-Norm statistics; we then use it to fine-tune the DNN, without any real data (CVPR2020).
4) Asynchronous training causes a degradation in generalization, even after training has converged to a steady-state.
We close this gap by adjusting hyper-parameters according to a theoretical framework aiming to maintain minima stability. (ICLR2020)