Видео

How to disabled edited tag symbol in what's app edit messages

Wilson Donald Perez Helen Taylor Sarah

Haha how are you Kane.

Thx!

great presentation!!

This is a very interesting talk -- quite helpful for my own research!

Is there an implementation of this algorithm (NTMA) anywhere for testing?

a natural question to ask is : can the random sparse pruning apply to GPT-style large models to achieve small models?

Thanks ! :D The LIFT idea is absolutely amazing !((:

Thanks for the great talk. I am trying to get a more inuitive understanding. In my mind's eye by taking a low-rank approximation of a network it is sort of like choosing the initialisation for the subsequent optimisation?

Wonderful talk.

lol🤣🤣🤣

*Promo SM*

from torch.optim import SGD from torch.optim.optimizer import Optimizer from torch.autograd import grad class SAM(Optimizer): def step(self, closure=None): loss = None if closure is not None: loss = closure() for group in self.param_groups: for p in group['params']: if p.grad is None: continue grad_params = grad(loss, p, create_graph=True) v = grad_params[0] p_s = -v / torch.norm(v) p.data.add_(p_s, alpha=lr) return loss # Usage model = ... optimizer = SAM(model.parameters(), lr=0.01) ############## # 2-step back prop # Backpropagate the image loss img_loss = mse(generated, real_fulldata.detach()) img_loss.backward(retain_graph=True) # Zero gradients manually before the first step of the SAM optimizer optimizer.zero_grad() # Perform the first step of the SAM optimizer (which is the base optimizer's step) optimizer.step() # Perform the second step of the SAM optimizer img_loss.backward()

Thanks!

Nice presentation on Robust RL, is there a link to have the slide please?

Nice talk!

Are the slides available somewhere? Thanks!

Thank you for the talk.

Awesome !!! Never ever thought sparsity in NN could be this interesting.

These talks are a really valuable resource, thank you for sharing them.

Informative session on Secure aggregation for federated learning and privacy concerns in multi-round secure aggregation.

so clear really thanks so much

😍😍😍

Was looking for few resources to understand what inductive bias meant. This video helped me a lot. Thank you

Thank you Maam, it was really a wonderful 70min of FL.

Thanks for giving such interesting facts. My question is in context of framework I.e. in which framework(TENSORFLOW/FATE etc) this work is carried out? And one more question is can we use this setting in analyzing traffic in ITS as there are different devices(Sensor,mobile devices,GPS, on-board sensor)which can participate in this?

The speaker was excellent but the question askers (Joseph started around 50:00 and Vijay, the host) were terrible and even malicious.

nice dedication 😉

the video image is too poor, you need to fix it more

Great talk!

how did you do it can you share with me , thank you

Can you please share the slides ?

awesome talk, very sexy approach on a hot topic!

A complete version of this work can be found at: arxiv.org/pdf/2105.10446.pdf