Thomas Bury

Doing maths in cardiac electrophysiology, climate science and ecology. Postdoc. McGill Uni.

[Thomas Bury @ThomasMBury]

Combining #AI and dynamical systems theory improves tipping point detection. Check out our new publication in PNAS: “Deep learning for early warning signals of tipping points”.
https://t.co/vLr0PhRi0S

Summary thread below:

[Thomas Bury @ThomasMBury]

Fantastic postdoc opportunity to work at the interface of maths and climate change:

https://t.co/0ySoJfLzXg

[Mehrshad Sadria @MehrshadSadria]

📢 Exciting News! 🧬 Introducing FateNet, a computational method co-developed by @ThomasMBury and I. By combining dynamical systems theory and deep learning, FateNet investigates cell decision-making processes using scRNA-seq data. 🧪 https://t.co/CGUZfUdhAQ

[Thomas Bury @ThomasMBury]

@thomas_quail @NatureComms @gilbublab @globalecochange @cbauch1 Thanks Thomas, and thank you for helping me with the chick heart data.

[Thomas Bury @ThomasMBury]

Bridging #AI and nonlinear dynamics can help us predict tipping points in discrete-time systems, notably heart rhythm! New paper out in @NatureComms with @gilbublab , @globalecochange , @cbauch1 and others:
https://t.co/Ny7ldw93DV

[Thomas Bury @ThomasMBury]

Excited to see the research that comes out of this centre!

[Sujith R I @SujithRI]

We have established a Centre of Excellence for studying Critical Transitions in Complex Systems @iitmadras. The Centre will perform research on predicting catastrophic events and developing early warning technologies for natural and engineering systems.

[Thomas Bury @ThomasMBury]

Listening to Spotify while training an ML model is a bad time...poor laptop.

[Yingjing Feng @SeffyYjFeng]

Greatly enjoyed #SIAMDS23 so far! 🤓 Check out our 3-part MS tmr & on Thur: “Phase Transitions in Electrophysiological Systems” (MS135, 155 &170) organized by @AravindKumar264 and myself from @SMQB_UoB. @TheSIAMNews

[Thomas Bury @ThomasMBury]

“The subtle math of a heartbeat gone wrong”, published in @PhysicsToday , is lucid description of our research with @khady_dgn, @gilbublab, and others. Thank you Johanna Miller for taking the time to digest and present our work with such clarity and wit!
https://t.co/NPFob6aVKt

[Thomas Bury @ThomasMBury]

Pleased to see my Python package ewstools published with JOSS. It provides tools to compute early warning signals for tipping points in time series data. Give it a spin - tutorials included. #OpenSource

[JOSS @JOSS_TheOJ]

Just published in JOSS: 'ewstools: A Python package for early warning signals of bifurcations in time series data' https://t.co/o9oR4Lte05

[Khady Diagne @khady_dgn]

Very happy to see my first publication, in @PhysRevLett . Wouldn't have been possible without our great team, congrats everyone!

@gilbublab @ThomasMBury

[Thomas Bury @ThomasMBury]

The mathematics of competing cardiac pacemakers is extraordinary. Really pleased to see this work with @khady_dgn, @gilbublab and others published in @PhysRevLett. Thanks to @philipcball for writing an excellent commentary.

[Gil Bub @gilbublab]

Our article on parasystole is now out in PRL, with a APS news commentary by Phillip Ball. Congats to Khady, Tom, Leon and the rest of the team! News: https://t.co/UOa1lDJFeZ
Article: https://t.co/xhg7y5tlLk

[Plotly @plotlygraphs]

#Plotly #Dash and #Python for heart health? ✅

@ThomasMBury used 𝘥𝘢𝘴𝘩.𝘥𝘦𝘱𝘦𝘯𝘥𝘦𝘯𝘤𝘪𝘦𝘴, 𝘱𝘭𝘰𝘵𝘭𝘺.𝘦𝘹𝘱𝘳𝘦𝘴𝘴 and callbacks to explore electrocardiogram recordings on Physionet as a (big) data source, supporting research on cardiac arrhythmias. 🫀

[Thomas Bury @ThomasMBury]

Interested in building a dashboard to interactively view your data? I've just written a @Medium article on how to do this using @plotly's Dash and Python. We use thousands of ECG recordings from Physionet as a use case:

https://t.co/fSssUYHjbY

[Thomas Bury @ThomasMBury]

Our deep learning model is trained on data that has been exclusively detrended using a Lowess filter. Using other filters (or no filter) can yield erroneous outcomes - please preprocess your data using a Lowess filter if using our model!

Thank you to @fdabl for catching this.

[Fabian Dablander (@fdabl.bsky.social) @fdabl]

Early warning signals for tipping points based on deep learning substantially outperform traditional indicators, as @ThomasMBury et al. showed.

In a short note, we illustrate an unintended behavior of the method, stressing the importance of preprocessing: https://t.co/IE5Odyz1Bb

[Sujith R I @SujithRI]

It gives me great pleasure to announce the launch of a webinar series on "Critical Transitions in Complex Systems" (https://t.co/GwvXg8ff3Z) jointly hosted by @iitmadras and @PIK_Climate . [1/9]

[University of Guelph News @UofGuelphNews]

#UofGResearch | Ecologist Dr. Madhur Anand (@globalecochange) and colleagues have created a model to account for how social polarization leads to inaction on fighting climate change.

@UofG @UofG_SES @UofGResearch
https://t.co/B1cq7HAbd8

[School of Environmental Sciences @UofG_SES]

New #SES_research published in @RSocPublishing by our own Dr. Madhur Anand (@globalecochange) in collaboration with @WaterlooMath shows how social systems, such as resource inequalities, influence #ClimateChange.

Original: https://t.co/T48QXpv01l

#Mitigation #UofG

[Thomas Bury @ThomasMBury]

@globalecochange @cbauch1 @TomBury1 @UofGResearch @UWaterloo @physorg_com @ThomasMBury <- This Tom Bury ;)

[Thomas Bury @ThomasMBury]

Thanks to all collaborators on this work: @cbauch1, @globalecochange, @MartenScheffer, R. I Sujith, Induja Pavithran & Timothy Lenton.

All code and data used in this study are available at https://t.co/1vQPM1bvEX and public repositories cited therein.

[Thomas Bury @ThomasMBury]

Combining #AI and dynamical systems theory improves tipping point detection. Check out our new publication in PNAS: “Deep learning for early warning signals of tipping points”.
https://t.co/vLr0PhRi0S

Summary thread below:

[Thomas Bury @ThomasMBury]

In conclusion: a deep learning classifier can be trained on data from a universe of possible models to detect and classify approaching bifurcations/tipping points with better performance than conventional early warning signals.