EPSRC Centre for Doctoral Training in Data Science. Training PhDs in machine learning, databases, stats+opt, & analysis of unstructured data at Edinburgh Uni
Edinburgh, Scotland
Joined November 2013
- Tweets1.1K
- Following265
- Followers1.1K
- Likes227
CDT in Data Science retweeted
Poster sessions are happening today! 11:40-1:20 and 5:00-5:30 today in room 208-210.
Come chat transfer learning for Gaussian processes and Bayesian optimisation 🌸
Excited to be in 🎊New Orleans🎊 this week for #NeurIPS2023! Ping me to chat Bayesian optimisation, transfer learning or ML and climate. Or if you know any good runs nearby!
And if you're interested in the above topics you should stop by our poster on Saturday!
CDT in Data Science retweeted
Do you love cross-lingual transfer?
Are you interested in putting latent variables everywhere you can?
Desperately searching for applied optimal transport research?
Come to my poster 16:00 in East Foyer! #EMNLP2023
virtual2023.emnlp.org/paper_…
w/ @tomhosking + Mirella Lapata
CDT in Data Science retweeted
🚨New TACL paper 🚨
Can we use explicit latent variable alignment for cross-lingual transfer? Minotaur uses optimal transport for explicit alignment across languages.
w/ @tomhosking, Mirella Lapata.
To be presented @emnlpmeeting + @mrl2023 in-person 🇸🇬
doi.org/10.1162/tacl_a_00611
CDT in Data Science retweeted
Excited for the poster session this afternoon at @automl_conf! I'll be talking about hyperparameter optimisation for increasing data set sizes and learning between tasks.
Work with Huibin Shen, @FrancoisAubet, David Salinas and @kleiaaro #AutoML23
Finding yourself repeatedly training your ML model because you’ve collected more data? Not sure what to do with the hyperparameters? We investigated for you!
Check out the paper arxiv.org/abs/2306.16916 with the wonderful Huibin Shen, @FrancoisAubet, David Salinas and @kleiaaro
CDT in Data Science retweeted
I'm really happy to share the news that Meta-Calibration has been accepted to TMLR! Meta-Calibration uses meta-learning as a new way to optimise for uncertainty calibration of neural networks. I've had a very positive experience with TMLR and certainly recommend submitting there!
Meta-Calibration: Learning of Model Calibration Using Differentiable Expected Calibration Error
Ondrej Bohdal, Yongxin Yang, Timothy Hospedales.
Action editor: Yingzhen Li.
openreview.net/forum?id=R2hU…
#calibration #prediction #optimise
CDT in Data Science retweeted
Life update: I have passed my PhD defense (@alexandrabirch1 and @licwu were excellent examiners), and moved to New York City!
Many thanks to everyone who made the PhD experience so fun, especially my CDT cohort, collaborators, and supervisor @driainmurray
CDT in Data Science retweeted
I'm very excited to be in Seoul to give a talk and present a poster about our CVPR'23 Meta Omnium paper at the Hyundai Vision Conference! Feel free to take a look at the poster and also a few photos from this great city! Meta Omnium website: edi-meta-learning.github.io/…
CDT in Data Science retweeted
Interested in what could be the long-term implications of deploying too many AI systems that are not fair? You can learn more in our paper - we've updated it on arXiv with the version that will appear soon in the proceedings of the Stanford Existential Risk Conference!
📢✨ New journal publication alert! ✨📢 @OBohdal, @tmh31 , Phil @OxfordTVG , and I @FazlBarez have dissected a crucial issue: lack of #AIFairness. Our analysis reveals how unfair AI can fuel social stress and unrest over time. 🧵 Here are the main insights: [1/4]
CDT in Data Science retweeted
Curious about how noise influences uncertainty calibration and generalisation of neural networks? We'll present our new work on this topic at the #ICML2023 SCIS workshop this Saturday! Collaboration with @MartinFerianc, @tmh31, @mrd_rodrigues
Hello from ICML'23! We will be presenting our joint work with @OBohdal, @tmh31, @mrd_rodrigues: Impact of Noise on Calibration and Generalisation of Neural Networks at the Spurious Correlations, Invariance, and Stability Workshop: arxiv.org/abs/2306.17630
CDT in Data Science retweeted
I'll be at #ACL2023 next week with 2 papers:
⭐️Extrinsic Evaluation of Machine Translation Metrics (Main conf - Monday Session 2: MT Oral)
⭐️Meta-Learning a Cross-lingual Manifold for Semantic Parsing (TACL - Wednesday Session 7: Semantics)
Reach out if you want to say hello!
CDT in Data Science retweeted
Finding yourself repeatedly training your ML model because you’ve collected more data? Not sure what to do with the hyperparameters? We investigated for you!
Check out the paper arxiv.org/abs/2306.16916 with the wonderful Huibin Shen, @FrancoisAubet, David Salinas and @kleiaaro
CDT in Data Science retweeted
Interested in multi-task and few-shot learning? Have a look at our recent @CVPR paper that introduces Meta Omnium!
Meta Omnium is a multi-task few-shot learning benchmark to evaluate generalization across CV tasks. Work by @OBohdal @EricTian1102, @yongshuozong, @chavhan_ruchika, @dali_academic, @henrygouk, @tmh31, will be presented on 20/6 afternoon. Project page: edi-meta-learning.github.io/…
CDT in Data Science retweeted
Interested in Bayesian optimisation and how it can be applied against climate change? Check out our new short survey. We look at material discovery, wind farm layout planning, optimal renewable control and environmental monitoring. For each we suggest a benchmark to get started.
CDT in Data Science retweeted
I’m very excited to come to #ICLR’23 in person! Together with @610v4nn1__ we’ll be presenting a poster for our paper “PASHA: Efficient HPO and NAS with Progressive Resource Allocation” tomorrow 1/5 at 16:30-18:30 CAT (MH1-2-3-4 #87). Looking forward to meeting some of you there!
CDT in Data Science retweeted
I’ll present our new paper "Fairness in AI and Its Long-Term Implications on Society" at the Stanford Existential Risks Conference today! The paper is also available on arXiv at arxiv.org/abs/2304.09826.
CDT in Data Science retweeted
I’m excited to share that our PASHA paper has been accepted to ICLR’23! PASHA allows you to quickly tune the hyperparameters and is already available in Syne Tune github.com/awslabs/syne-tune…
If you need to tune the hyperparameters (or the architecture structure) of you ML algo but you don't have much time, take a look at our ICLR2023 paper: openreview.net/forum?id=syfg…
@OBohdal
@lukas_balles
@cedapprox
CDT in Data Science retweeted
Our @EDINrobotics CDT program is now recruiting PhD students for September 2023. If you would like to work with me and the Interaction Lab (sites.google.com/site/hwinte…) on grounded language learning, Embodied AI, and multimodal learning, please reach out! #NLProc #dlearn #Robotics
CDT in Data Science retweeted
Check out our new preprint (w/ @nikita_moghe @alexandrabirch1 and Mark Steedman) on how segment-level MT metrics can't predict downstream cross-lingual outcomes!
🚨 New preprint with @tomsherborne, @alexandrabirch1, and Mark Steedman
We evaluate segment level MT metrics on extrinsic tasks and they show a negligible correlation 😮
arxiv.org/abs/2212.10297 [1/5]
CDT in Data Science retweeted
Just one more day to go! Tomorrow’s the day for our @NeurIPSConf social on ML and climate change! Join us in room 392 at 6pm for networking, bingo and roundtable discussions. And join our slack too join.slack.com/t/climate-ml-… !
CDT in Data Science retweeted
Just one week till our @NeurIPSConf social on ML and climate change! We’ll have icebreaker bingo, roundtable discussions and a friendly atmosphere. sighellan.github.io/climate-…