I'll be speaking on TDA, angiogenesis, and disease diagnostics today at #SIAMLS22 in MS18 at room 336 from 4:30-5:00. meetings.siam.org/sess/dsp_p… Come check it out and say hi!

4/4 ... But, in case study 4, we get great parameter estimation using a modified logistic model w/ spatial correlations from bit.ly/3yoPE1P. But, these types of models are difficult to derive in practice.

2/4 Re:AIC/BIC: In previous work in #BullMathBio (w/ simulated PDE data), we did find situations of overfitting w/ complex learned models. To remedy, we learned 3 models and then used AIC on simulated models for selection (step 3 of pipeline). bit.ly/3jfOcIs

The good part of not being able to sleep at 4 AM is catching @a13xbrowning 's Paper prize winning talk @smb2021! Congratulations and well deserved, Alex! Fantastic talk.

...to partition many model simulations into robust and biologically interpretable groupings based on input parameters! Joint work with Bernadette Stolz, @haharrington, Kevin Flores, and Helen Byrne Pre-print: tinyurl.com/22k4va4h Code: github.com/johnnardini/Angio… 4/4

Unfortunately, this discrete agent-based model is difficult to study in a data-driven manner due to its stochastic and discrete nature. I'll be discussing how topological data analysis methods can overcome these challenges ... 3/4

The Anderson-Chaplain model (tinyurl.com/3y6dj4h9) of tumor-induced angiogenesis is beloved in the Mathbio community because it is simple yet able to capture experimentally-observed blood vessel morphologies 2/4

I'll speaking at @smb2021 today in the CDEV-1 MS session (9:30-11 AM Pacific /12:30-2 PM Eastern/5:30-7 PM BST/ 1:30-2 AM KST) on "Topology discriminates parameter regimes in a model of angiogenesis" @SMBdevBio 1/4

Curious to see what all the buzz is about ...

Sunday granola 😋😋

But sometimes these models give misleading predictions – then what do we do? Our tutorial discusses how equation learning (EQL) can help us learn other DE models to predict ABM output, and investigates the robustness of this methodology to many data challenges. 4/5

Fortunately, simple ABM rules can be coarse-grained into differential equation (DE) models, such as mean-field models, to predict ABM behavior. Here is a birth-death-migration model 3/5

Where is everyone?

Please RT. Happy to announce our MS "Leveraging Machine Learning for Discovery of Mathematical Models in Biology" for @TheSIAMNews MDS will be held online! June 2, 3-5 PM EDT. Title and abstracts: bit.ly/2Zr5om2 Registration: bit.ly/2zpvBab

And don’t forget this classic

Beyond proud of @ahduprey , Faneul Sisay, Natasha Stewart, and Yangxinyu Xie!! My undergrad equation learners have been working hard at the @SAMSI_Info undergrad modeling workshop and gave an AWESOME presentation on the performance of LASSO with the SINDy algorithm!!!

Happy holi!

Had a great week at the #amsmrc18

Maybe one day...