Sinead Williamson
@sineadwilliamso
Followers
2,108
Following
448
Media
10
Statuses
415
Machine Learning/Statistics. MLR Apple / Assoc. Prof UT Austin. Mostly Bayesian. She/her. Opinions my own.
Seattle, WA
Joined July 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Texas
• 285633 Tweets
Yankees
• 251617 Tweets
WANG YIBO IN THE GT FINALS
• 155207 Tweets
Presiden
• 128396 Tweets
World Series
• 96947 Tweets
Juan Soto
• 65687 Tweets
期日前投票
• 60615 Tweets
HAPPY TWENTY FOURTH
• 47365 Tweets
#菊花賞
• 41510 Tweets
アーバンシック
• 39532 Tweets
ダノンデサイル
• 32424 Tweets
gerard
• 22702 Tweets
Happy Mashujaa
• 21235 Tweets
#虹7th_day2
• 21001 Tweets
Gowon
• 19710 Tweets
りんご飴
• 19348 Tweets
ダービー
• 16200 Tweets
#スパスタ3期
• 15094 Tweets
ねこあつめ2
• 14902 Tweets
メイショウタバル
• 14536 Tweets
コスモキュランダ
• 14465 Tweets
My husband is traveling with me to a conference and we’ve been thinking of fake AI subfields for him to claim to be working on. Favorites so far:
- Disentangled HairNets
- Deep Distraction Networks
- Misrepresentation Learning
Got any other good suggestions for him?
465
372
3K
Just received what might be my favorite ever line in a review: "The idea is not too complicated, but it is neat and worth while. It is one I wish I had." 🥰
7
13
589
📢Internships at Apple ML Research🍏
We’re looking for a PhD research intern with interests in uncertainty quantification, LLMs, probabilistic ML and/or decision making under uncertainty!
See thread for more details 👇
[1/3]
4
73
505
Oh, great. Neurips desk reject. If the results are “known results” (I’m pretty confident they aren’t), perhaps the area chair could have made the effort to leave a note telling what they are? 1/n
14
19
307
🎉I found out on Friday that I was approved for promotion and tenure in
@UT_Stats
🎉 - huge thanks to the mentors, students, collaborators, colleagues and support staff who made this possible 🥰
27
1
282
We’ve got a long way to go, but this year’
#NeurIPS
has felt more inclusive and self aware and has been my favorite yet. Kudos to
@InclusionInML
,
@black_in_ai
#QueerInAI
@_LXAI
@WiMLworkshop
and all the fantastic workshops promoting fairness and social engagement.
1
23
123
I am looking for interns at MLR Apple interested in Bayesian ML/DL. If you're interested please apply here! (the description in the link isn't team-specific and fairly generic, but don't worry - you're in the right place)
5
22
97
Disclosure: I am an area chair for neurips. I desk rejected two papers. I left them as long a review as the space allowed, including citations of the work that is similar. Feels like common courtesy.
3
3
91
This Friday at
@NeurIPSConf
, I'll be hosting a
@WiMLworkshop
mentoring table on navigating the academic job search and tenure application procedure. Come chat with me! Or, check out one of the other great mentoring tables at
0
17
73
Disappointing to have work rejected with a pro forma check box response. Got to be even more disappointing for newer people. 3/3
1
3
70
When creating figures for papers, remember that over 4% of your audience are colorblind and many more will be using black and white printers. Choose color schemes that make it easy for them! I like Paul Tol’s schemes and analysis:
3
15
63
@ylecun
And yet people use this dataset. If I buy moldy bread and don’t inspect it before selling sandwiches I made using it, I don’t get to pass all blame back to the bakery.
1
1
54
👀 On the faculty job market in Statistics, Biostatistics, Machine Learning, or similar? UT Austin is hiring!
Got questions? Ask me! (email in link) Got students or post docs who might be a good fit? Tell them!
0
18
50
You don’t even need to be registered for the conference!
@sineadwilliamso
and I will be moderating the
@NeurIPSConf
tutorial by Wee Sun Lee (NUS) on "Message Passing in Machine Learning". Video is free to watch for everybody. Post your questions at Rocketchat if registered, or even ask them here if u aren't.
Hope to see you there.
1
11
82
0
4
48
Why aren’t the default color palettes in matplotlib and seaborn colorblind-friendly? At least seaborn has a named ‘colorblind’ palette, but why not make the most accessible palette the default - to my non-colorblind eyes it is just as attractive and interpretable as the default
4
3
45
Interested?
First email us at mlr-uq4llms-internship
@group
.apple.com including your CV and highlighting your most relevant experience, and then apply at .
[3/3]
1
2
45
The problems with ML peer review seem to boil down to two issues: artificial scarsity and lack of reviewers. Not sure how to deal with the latter. But the new online world, and the prospect of partially online future conferences, suggests solutions to the former 1/n
1
4
41
@FromPhDtoLife
@wimlds
Took up roller derby. Great for my physical and mental health, plus meant that when I moved for postdocs and faculty positions I was able to make friends via new teams. I don’t think I could have taken so many solo moves otherwise.
0
0
31
Excited to be moderating this awesome WiML panel discussion at ICLR, centered on research about, and research during, COVID-19. Hope to see you Thursday! Join the wiml group on ICLR rocket chat for more details
Excited to announce the panelists of our Virtual Social @
#ICLR2020
on Apr. 30th 9am PST:
- Sasha Luccioni
@SashaMTL
- Lily Peng
@lhpeng
- Morine Amutorine
@theAmutorine
- Cecilia Mascolo
@cecim
- Katherine Heller
@KatherineHeller
Moderator: Sinead Williamson
@sineadwilliamso
1/3
1
16
46
1
5
30
I’m excited to talk at the
#ICML2020
#NewInML
workshop on Monday about mentorship and building a supportive community as a new researcher. Are you currently involved in a machine learning
#mentoring
program or event? Let me know, I’d love to amplify it!
0
9
26
🎓This is a pure-research internship - the goal is to publish new and interesting ideas
✈️The internship will be based in Seattle; we will support necessary visa/travel arrangements
📅Internship duration 4-6 months, any time between January-September 2025.
[2/3]
1
2
28
Excited to be at the Bayesian nonparametrics conference in Oxford. I’ll be taking about scalable inference later today, and my former student
@zhang_michael91
will be talking about our work on dynamic latent feature models on Thursday.
1
2
24
Very proud of this paper with Michael Zhang
@zhang_michael91
, Avinava Dubey and Eric Xing that describes a very general distributed MCMC algorithm that’s appropriate for almost all Bayesian nonparametric models. 1/2
1
0
23
UT Austin Statistics is hiring at all levels. Come be my colleague!
💻 We’re hiring! Three faculty positions starting Fall 2021.
Learn more:
0
34
30
0
5
23
Why would this be caused “doublind” when it’s anything but double blind? Best case, this becomes a way to fawn over popular researchers and leave anonymous self-reviews, worst-case it’s yet another way to harass subsets of the community.
Finally! After all these years dreaming this up, someone put together a social peer review platform that lets you upvote papers and leave comments. This + arxiv 👌. (Ok, the challenge will be to detect fake comments and votes, but at least it's a start.)
16
51
365
2
1
23
Requiring in-person attendance disproportionally impacts researchers from countries that require visas, researchers with limited financial resources, and researchers in remote locations.
„Data science and machine learning conferences (NeurIPS, ICML, AISTATS, ICLR, UAI, ...): Allow remote paper & poster presentations at conferences” -
7
173
528
1
4
21
Most delicious talk of
#WiML2019
so far...
@XandaSchofield
on the relationship between topic models and
#GBBO
1
3
20
Looking forward to talking about Bayesian Nonparametrics at the SMILES summer school later today - thrilled to be part of such a great event (and with so many great Bayesians!)
There are other Bayesians giving tutorials there as well (
@andrewgwils
, Tamara Broderick,
@sineadwilliamso
), and many other amazing speakers too. Check it out:
Thanks to
@burnaevevgeny
@RodrigoRivr
@SMILESkoltech
for organizing.
0
7
35
0
2
20
@danielmgmi
@moarbugs
@eredmil1
There are enough barriers to entry for scientists from Africa, Latin America and the Middle East already. We should be removing those barriers, not erecting more.
0
3
19
@NeurIPSConf
I think this is the main thing I’d like to see resolved: area chairs should have been required to leave a response. If you’ve read enough to reject, you’ve read enough to write a short review - the time sink is the reading and deciding, not the writing 1/2
@GuoOctavia
@sineadwilliamso
The design was bad imo: there were boxes to check, each corresponding to one sentence like this, and then an “other reasons” box that let you give more details. I also put in max-allowed-length “other reasons” for everything I desk rejected; agreed that it’s common courtesy.
1
0
11
1
3
18
@lzamparo
If any of these end up in NeurIPS this year, they better include my husband as a co-author.
2
0
19
The
#WiML
Un-Workshop at
#ICML2020
features a panel discussion July 13 at 19:35–20:35 GMT. Excited to introduce our panelists:
- Doina Precup
- Deborah Raji
@rajiinio
- Anima Anandkumar
@AnimaAnandkumar
- Angjoo Kanazawa
@akanazawa
Moderator: Sinead Williamson
@sineadwilliamso
0
23
90
0
2
19
Currently we evaluate “is this in the top x% of submissions” for small x, not “is this good enough to publish”. This encourages a mindset of looking for reasons to reject, not looking for reasons to accept. 2/n
2
0
15
@dmimno
“A random elephant is just some function from \Omega to a suitable space of elephants.” From Kingman’s Poisson Processes, my favorite textbook of all time.
2
0
15
As a Texan this is the exact opposite of my opinion on both matters
Perry says Texans wiling to suffer blackouts to keep feds out of power market
4K
2K
4K
0
0
15
About time. They’ve known about these problems for years thanks to
@jovialjoy
@timnitGebru
@AJLUnited
and have minimized and deflected.
Amazon imposes one-year moratorium on police use of its facial recognition technology by
@kharijohnson
0
11
21
0
1
14
People who step in to do last minute conference reviews, after the assigned reviewer ghosts, are the best sort of people ❤️. So glad so many people were willing to step up and do last minute
#aistats
reviews for me; so annoyed that they had to.
1
0
14
Super excited to join
@WiMLworkshop
and
@cwstat
for this event at
@aistats_conf
tomorrow! Take part in community-driven mentoring sessions then join us for a panel discussion on publishing! 1/2
1/
#WiML
is excited to host a joint event with the Caucus for Women in Statistics (
@cwstat
) at
@aistats_conf
. The event has two components: community-driven mentoring and a panel, and it will be held on the Icebreaker video platform on Tuesday, April 13, 12.30–2pm PT.
#AISTATS2021
1
8
17
1
3
14
Kicking off
#WiML2019
with a great talk by
@dawnsongtweets
on Security and AI
#wiml
#NeurIPS
#NeurIPS2019
@WiMLworkshop
0
4
12
I know that you, and the field, are in a bind: so many papers, not enough qualified reviewers. I don’t love early rejects, but I understand the reason for trialing it. It just feels it could have been so much kinder and more helpful, for little extra work. 2/2
1
1
12
@hoonoseme
@NisheethVishnoi
@profelisacelis
@InclusionInML
@black_in_ai
@_LXAI
@WiMLworkshop
Having Diversity and Inclusion Chairs sends out the message that this is a serious issue and provides mechanisms for change.
@haldaume3
and
@kat_heller
did a fantastic job at this role, and I hope to see all conferences having such a position.
2
3
12
I have a mirror. I know I look my age. When people say “you look too young to be a professor” they mean “you look too junior to be a professor”
No, I will not “take it as a compliment” when someone asks me if I’m a grad student
I’m 42 years old
I’ve worked hard to be where I am
I’ve overcome discrimination
I’ve overcome harassment
I don’t want to be praised for my appearance
I want to be respected for my expertise
237
826
14K
0
1
11
Just watched
@CodedBias
with
@jovialjoy
@rajiinio
@mathbabedotorg
@safiyanoble
@shalinikantayya
and many others. SO GOOD. So important. Watch it. Get people at your AI-using/building/selling company to watch it.
0
0
11
Opportunity Alert! Women in Machine Learning (WiML
@WiMLworkshop
) is recruiting new board members for its Policy and Research Committee. For more details and to apply, visit
0
6
10
Congratulations to Dr
@elahe_ghalebi
on a successful thesis defense! Her thesis on graph-based models covers Bayesian nonparametric models for dynamic and partially observed graphs, and compressive sensing methods for graphs. It’s been an honor to co-supervise you!
1
2
11
The number of times I’ve seen people model what is essentially linear data with a neural network is astounding.
unknown to most outside the field, DL often does not outperform far simpler models. With obvious exceptions for NLP, vision, and RL.
9
41
208
1
0
10
@skoularidou
You are such a strong and inspiring person and this is so very, very unfair. I really hope you prove him wrong.
1
0
10
I’ve been seeing a lot of tone policing of black women in AI lately, who dare to suggest our research is harming them. Strangely, didn’t see similar tone policing a few months back when a bunch of white folk were busy getting worked up about connectionism vs symbolism. Stop it.
0
0
10
Looking forward to talking soon with
@oxcsml
about work on interaction networks with coauthors
@elahe_ghalebi
@HamidrezaMahyar
Graham Taylor and Radu Grosu!
Join us tomorrow at 3.30pm GMT for our next OxCSML seminar by
@sineadwilliamso
who will be talking about 'Bayesian nonparametric models for interaction networks'!
Zoom details available following registration at .
0
2
14
0
0
10
Check out our new JMLR paper on parallel inference for Gaussian Processes, with
@zhang_michael91
0
0
9
Excited for this year’s WiML workshop - what a stellar line-up 🤩
The
#WiML2019
program features 8 remarkable contributed talks by: Kimia Nadjahi (
@TelecomParis_
), Xinyi Chen (
@Google
), Liyan Chen (
@FollowStevens
),
@snehaark
(
@Google
), Qian Huang (
@Cornell
), Mansi Gupta (
@PetuumInc
), Margarita Boyarskaya (
@nyuniversity
),
@natashajaques
(
@MIT
)
1
19
56
0
0
9
If you're at
#NeurIPS2020
, check out Mridula Maddukuri
@Summer1992
and my poster on Deep Multi-Instance/Multi-Label Learning at the Joint Affinity Group poster session today and the
@WiMLworkshop
poster session on Wednesday!
0
3
9
Hey
@Kings_College
- I chose to apply to you in part because of your great record on state school admissions. Really hoping to see a statement similar to
@WorcCollegeOx
that you are honoring offers to (disproportionately state school) students affected by the A-level debacle.
1
0
9
@St_Catz
Another alum hoping you reconsider. It seems
@WorcCollegeOx
are managing to stretch their numbers.
0
0
8
@dwf
@NeurIPSConf
Even when I don’t think a paper is likely to get in based on reviews, I still find writing a rebuttal is useful for me as an author. It forces me to really engage with the critiques which can make the next version better.
1
0
8
Top x% is needed when we have a limited number of talk slots or poster capacity. But recent conferences have shown that we can be much more creative than the old model, and still highlight the best papers, suggesting it is feasible to move to a quality based bar not a % bar. 3/n
1
0
8
Great advice for those presenting at
#ICML2020
this week!
0
1
8
I used to have a HW question on spurious correlations/adjusted R-sq based on “I wish” by Skee-Lo. It had a fake dataset, with columns for height, baseball skills, ownership of rabbits, hats, and 64 Impalas. No-one got the reference. Or maybe they were just embarrassed for me.
1
1
7
@adjiboussodieng
@IBJIYONGI
Inclusion should firstly benefit the individual. Lifting up individuals is reason enough, regardless of the benefits to the <field/department/company> from diversity (which can be huge) 2/2.
0
2
7
In fact, this should have a knock-on effect on the # of reviewers problem, by reducing the number of times papers are submitted, recieve weak accepts, but are rejected - with each submission taking up 3+ reviewer slots. 4/4
0
0
7
@thegautamkamath
Getting to meet senior people who are interested in what you are doing was definitely one of the (nervewracking!) highlights of my early conferences! We hope this allows that sort of interaction to thrive in a virtual environment.
0
1
7
@daniel_bilar
@bschoelkopf
@IgorCarron
Allowing remote attendence doesn’t necessitate banning in-person attendance.
1
0
7
@hoonoseme
@NisheethVishnoi
@profelisacelis
@InclusionInML
@black_in_ai
@_LXAI
@WiMLworkshop
Increased accessibility.
@black_in_ai
gave funding to many students from underrepresented countries, and were thoughtful about how they did so (paying for expenses up front rather than offering reimbursements).
@WiMLworkshop
funded hundreds of female attendees, often first-timers
1
1
7
Congratulations to
@guywcole
for a successful PhD defense! It’s been great working with him at
@UT_Stats
on fairness and relational models.
@OfficialOpcity
are lucky to have him!
0
0
7
@roydanroy
Open book, randomized question versions so no two students will have the same exam. Upload handwork within 10 mins of submitting. Increase number of exams but decrease weight to reduce the urge to cheat on any given exam.
0
0
7
0
2
7
Looking forward to taking part in the I Can't Believe It's Not Better workshop later today! Many thanks to
@AaronSchein
,
@MelaniePradier
,
@in4dmatics
,
@franciscuto
,
@FinaleDoshi
,
@hannawallach
,
@blei_lab
and Isabel Valera for orchestrating such a cool & thoughtful workshop.
Last and definitely not least will be a PANEL moderated by
@hannawallach
. Esteemed panelists are
@sineadwilliamso
,
@ta_broderick
,
@laurent_dinh
,
@KLdivergence
, and
@lawrennd
.
10/11
1
2
9
0
1
7
Well, shit. Sorry y'all.
@HowellONeill
@mims
this is the one for me, too. Can you imagine a british person saying "y'all". It'd be appalling
2
0
6
0
0
7
Congratulations to Maurice Diesendruck
@momodomo
on a successful thesis defense! It’s been great working with Mo on synthetic data and privacy.
0
0
7
Today marks a new semester and a new administration. Excited for both.
0
0
7
“This seems cool”
+
“Privilege”
The latter let me follow the former
Tell me why you went into your field in 5 words or less.
20K
690
12K
0
0
7
Yup, couldn’t find my childhood vaccination records so had to get re-jabbed for almost all of these. Cost several hundred dollars.
FYI, all immigrants to the US are *required* to have these vaccines:
Mumps
Measles
Rubella
Polio
Tetanus and diphtheria
Pertussis
Haemophilus influenzae type B (Hib)
Hepatitis A
Hepatitis B
Rotavirus
Meningococcal disease
Varicella
Pneumococcal disease
Seasonal influenza
14
60
322
0
0
6
"What you first need understand is that 30-35 years ago we 100% considered ourselves to be in modern, liberal times, absolutely unbiased, and completely fair." Great tweet thread on bias and hiring
I wrote this thread a long time ago but decided not to put it up. Today, because of other things on twitter that you will recognize, I have dusted it off. It is a little raw.
6
211
957
1
1
6
Super excited to be a part of this workshop at
@netsci2019
!
Announcing the 6th Statistical Inference for Network Models (SINM) workshop
@netsci2019
! With 4 amazing invited speakers:
@sineadwilliamso
@ta_broderick
@austinbenson
@Eliz_Hobson
. Co-org with
@DanLarremore
@baileyfosdick
@tinaeliassi
. Deadline March 11:
0
30
65
0
2
6
@le_roux_nicolas
@doomie
I hope we don’t have in-person unless ALL countries are vaccinated, not just rich ones. I can picture an amplification of previous years’ visa issues.
0
0
6
So much of what I know about the Civil Rights movement in the USA, and about my current, meagre understandings of race in modern America, started with reading the March Trilogy. Thank you, John Lewis. RIP.
Awful, awful news: Civil Rights Icon Rep John Lewis has died.
2K
10K
64K
0
0
6
Algorithic Justice League
@AJLUnited
, Black Girls Code, Code2040, Deep Learning Indaba
@DeepIndaba
, any of the ML affinity groups (
@black_in_ai
,
@_LXAI
,
@QueerinAI
,
@WiMLworkshop
)
2
0
6
Looking forward to
@cwstat
's round table on "Journeys in AI, ML and Stats: the female perspective" -- starting in 30 mins at
@aistats_conf
!
0
0
6
@roydanroy
This is why you should always sleep on a review before submitting it. I often realize my first draft was harsher than it should have been because I was over-focused on one point.
0
1
6
1
0
5
New
#uai2019
paper on sparse random graphs with dense subgraphs - check out the paper here: or stop by the poster session tonight if you’re in town!
0
2
5
@hoonoseme
@NisheethVishnoi
@profelisacelis
@InclusionInML
@black_in_ai
@_LXAI
@WiMLworkshop
Yes! Some concrete things: Use of, and promotion of, inclusionary language. The conference name is a big one, but the pronoun stickers are a smaller example. Branding and visual signals are important and send a constant, low-level reminder of what is and is not acceptable.
1
1
5
@ampanmdagaba
@roydanroy
@rasbt
Well, the research suggests that single blind reviews disproportionately benefit famous authors and top universities
4
1
5
@seanpmackinnon
@hellpumpking
@ChelseaParlett
Oh wait I ended too soon. Want heavier tails? Throw on a gamma prior. Want literally any discrete distribution on the nonnegative integers with full support? You can express it as a mixture of Poissons.
1
0
4
@csdoctorsister
@math_rachel
@cfiesler
Agree and I wish it had been called that from the beginning (in part because, call it what it is! But also because bias already has a very specific meaning in statistics and ml, and the overloading causes confusion and ambiguity)
0
0
5
Really disappointed by the way Google has treated
@timnitGebru
and how the field continues to treat Black women who question the status quo.
#ISupportTimnit
#BelieveBlackWomen
It's such a shame how
#AIEthics
has been entirely dependent on women of color for their situated knowledge and resilience yet treats their talent as disposable when they do the hard work of challenging institutions.
#ISupportTimnit
0
0
11
0
0
4
@BasicScienceSav
@AcademicChatter
Absolutely ask your direct advisors! It is almost a red flag not to have them write a letter - absent other information it implies they declined
0
0
4
@hoonoseme
@NisheethVishnoi
@profelisacelis
@InclusionInML
@black_in_ai
@_LXAI
@WiMLworkshop
@haldaume3
@kat_heller
Complimenting that, support and involvement of established names in the field. Having people the community look up to -eg
@JeffDean
@AnimaAnandkumar
and many more - stand up for change makes a huge difference in terms of mass behavior.
1
0
4