![Vitaly πΊπ¦ Feldman Profile](https://pbs.twimg.com/profile_images/1347743048042639365/_IpmgfTD_x96.jpg)
Vitaly πΊπ¦ Feldman
@vitalyFM
Followers
4K
Following
2K
Statuses
330
Proving theorems about machine learning. Worried about data privacy. Research scientist at @Apple
Mountain View, CA
Joined August 2016
RT @gingsmith: Are you a junior researcher interested in federated and collaborative learning? Consider applying to the Simons Institute Reβ¦
0
12
0
@QuanquanGu Even bad prior experiences don't justify biases against entire nationalities. My personal experience and a documented fact is that a disproportionate amount of dishonest research behavior originates in China. But everyone deserves to be treated fairly based on their own merit
0
0
9
RT @TheGradient: Workshop on Theory & Practice of Foundation Models (organized by @mirrokni and myself) will happen this week @GoogleAI Mouβ¦
0
23
0
Strange to see such misguided article reposted by @sapinker . The most obvious problem with the analogy is that grid operators do not provide cheap energy when it's "free" due to solar. There is plenty of flexible electricity demand for it (e.g. charging EVs and training LLMs)
0
0
3
Nice work! It's not a paradox for a while though. The trade-off has been observed in several closely related contexts (model compression and privacy) and explained in my work on the role of memorization for long tail distributions
𧡠1/ We use weight decay everywhere. Itβs a go-to for improving generalization and stabilizing training, right? But hereβs the catch: it can also make models give up on low-frequency classes (π±). Not ideal!
0
1
10
@thegautamkamath This warning isn't going to deter anyone considering collusion and to me sounds more of an admission that NeurIPS still has no real way to deal with major collusion problems at ML conferences
2
1
17
RT @davidmarcus: Went to UCLA this afternoon to talk to people peacefully. Few interesting takeaways: 1) Most of them openly say they suppβ¦
0
5K
0
@inhaleopenair @Mathgarden This troll is desperately trying to turn an important and factual education policy discussion into personal attacks on the integrity of people like you and @minilek . Clearly he has no idea who he is dealing with and found a great way to discredit himself.
3
0
8
RT @Grananqvist: I'm excited to announce that we have open sourced pfl-research β a fast, modular and easytouse Python framework for simulaβ¦
0
6
0
@oveisgharan Try as you may, it's not up to you (or Al-Jazeera style propaganda) to define genocide. Even the Nazis weren't using their population as human shields in dense areas. The reality is that even the current ratio requires major sacrifices from Israel (200+ dead soldiers).
0
0
0
@oveisgharan To be clear, I'm not saying that the innocent lives don't matter. They certainly do and the high number definitely raises concerns about whether the rules of war and humanitarian principles are being ignored. Labeling the whole thing as genocide is not going to help find this out
0
0
1
RT @Billbrowder: Apple agrees to pay $12.3m to the Russian budget, which will now be used to kill Ukrainians. Why is Apple still doing busiβ¦
0
2K
0
@mahdi_tcs The only way to make such claims is not not what you are talking about. The reality is that Hamas is an actual and brutal threat to the entire region and talk of restraint is cheap when it comes from the safety and comfort of not being targeted daily
0
0
0
Automatic subtitling of talks has become a standard feature at big ML confs @NeurIPSConf. It fails often even for clear and native speakers so I can't imagine anyone being able to rely on it. But at least it's rather entertaining. Here is one example and feel free to add yours
1
0
14
@mahdi_tcs And before accusing Israel of unheard brutality toward (the rather hostile) Gazan civilians it's worth reviewing how Assad's regime (and Russian army) treated its own people in the war with the rebels.
0
0
1