Thursday, July 4, 2024

NIH funding

Bridge to Artificial Intelligence (Bridge2AI) Program
- Description: This NIH Common Fund program aims to propel biomedical research by generating new data sets and best practices for AI and machine learning analysis. The initiative is designed to address complex biomedical challenges beyond human intuition.
- More Information: Visit the Bridge2AI program page for detailed information and updates (NIH Common Fund).
Smart Health and Biomedical Research in the Era of Artificial Intelligence (AI) and Advanced Data Science (NSF 23-614)
- Collaborators: National Science Foundation (NSF) and NIH.
- Objective: Supports projects that integrate AI and advanced data science to enhance biomedical research, focusing on innovative health solutions.
- Submission Deadline: November 13, 2024.
- More Information: Details can be found on the NIAID funding news page (NIAID).
Transformative Artificial Intelligence and Machine Learning Based Strategies (R21/R33 Clinical Trial Not Allowed)
- Objective: Develop strategies to identify determinants of exceptional health and lifespan using AI and machine learning.
- Status: No current funding opportunities, but it represents the kind of initiatives NIH has supported in the past.
- More Information: Visit the NIH Common Fund page for archived initiatives and future updates (NIH Common Fund).

For updated deadlines, detailed descriptions, and additional opportunities, regularly check the NIH Grants & Funding page.

der to understand human health and disease etiology. This includes multi-omic data acquisition undertaken by ongoing exceptional longevity (EL) studies supported by the National Institute on Aging (NIA) that aim to identify and translate protective molecular factors and biological processes that promote exceptional health and life span. Such NIA-supported studies include the Long Life Family Study (LLFS), Longevity Consortium (LC), and Integrative Longevity Omics (ILO). While these human cohorts, with extensive physiologic, clinical, and pharmacologic data, provide advantages to unravel exceptional aging processes, the limited signal strength caused by modest variance in life span across humans and other stochastic factors, such as environmental exposures, could hinder the detection of protective biological factors that drive EL. In an effort to overcom

Center for Equitable AI and Machine Learning Systems

https://equitableaisymposium.com/#speakers

Wednesday, July 3, 2024

China leading generative AI patents race, UN report says

https://www.reuters.com/technology/artificial-intelligence/china-leading-generative-ai-patents-race-un-report-says-2024-07-03/

DPGR and policy

Tuesday, July 2, 2024

ACTION AI Institute has developed the MABEL dataset

The ACTION AI Institute has developed the MABEL dataset, which stands for Malware Analysis Benchmark for Efficient Artificial Intelligence Modeling and Machine Learning. This dataset is designed to aid in the development and benchmarking of AI models for malware analysis. The dataset and related resources are hosted on GitHub, where you can find the repository containing the dataset, code, and documentation.

Trustworthy ML

https://www.trustworthyml.org/symposium

https://www.youtube.com/playlist?list=PLNfU-a7sxIwvS7dhnOPdFtvhdNcrnufEW

Monday, July 1, 2024

ODU CS PhD application https://www.odu.edu/computer-science/academics/graduate/phd/admissions

https://www.odu.edu/computer-science/academics/graduate/phd/admissions

Home Computer Science Academics Graduate Ph.D. in Computer Science PhD Admissions

PhD Admissions

Overview

First, you should view all the information and admissions requirements at ODU Graduate Admissions and ODU International Admissions, if applicable, including ODU's English proficiency requirement.

Deadlines

Location	Fall	Spring	Summer
Domestic	June 1	November 1	March 1
International	April 15	October 1	February 1

Showing 1 to 2 of 2 entries

Requirements

GRE
Recommendations from 2 faculty or employment supervisors
Transcripts from all prior institutions

Students are expected to show proficiency in Problem Solving & Programming, Introduction to Computer Architecture, Advanced Data Structures and Algorithms, Introduction to Theoretical Computer Science, and Operating Systems at an undergraduate level.

Students may be admitted directly to the PhD program with either a bachelor's or a master's degree, preferably with a computer science related major, but that is not a requirement. However, if you do not have a BS in Computer Science, you may be required to complete undergraduate prerequisites before you are eligible to take graduate courses or hold a Graduate Teaching Assistantship.

Applying

When you are ready, apply for admission online. Once the application is complete, the graduate committee will evaluate the application. The committee looks at the whole application before making a decision. In general, the committee considers three factors for admitting students into our PhD program:

GPA and ranking of the graduating school
GRE scores (there is no cutoff for considering the application)
MOST IMPORTANT: Student background in terms of educational training and research interest should match with one of our faculty. For this you should browse the CS Faculty Directory and Faculty Disciplines. You are encouraged to contact the faculty directly to gauge common research interests. (If you have not contacted a faculty member to discuss common research interests, your chances of admission are much lower.)

All full-time PhD students receive a 100% tuition waiver and a stipend.

If you have further questions, please contact the Graduate Program Director (GPD) for Admissions.

COMPUTER SCIENCE

APPLY NOW

CONTACT

DEPARTMENT OF COMPUTER SCIENCE

3300 Engineering & Computational Sciences Building
Norfolk, VA 23529

757-683-7740

office

csgpd@odu.edu

QUICK LINKS

5115 Hampton Blvd
Norfolk, VA 23529

Resources

Legal

Visit

Choose LanguageLabelAfrikaansAlbanianAmharicArabicArmenianAzerbaijaniBasqueBelarusianBengaliBosnianBulgarianCatalanCebuanoChichewaChinese (Simplified)Chinese (Traditional)CorsicanCroatianCzechDanishDutchEnglishEsperantoEstonianFilipinoFinnishFrenchFrisianGalicianGeorgianGermanGreekGujaratiHaitian CreoleHausaHawaiianHebrewHindiHmongHungarianIcelandicIgboIndonesianIrishItalianJapaneseJavaneseKannadaKazakhKhmerKoreanKurdish (Kurmanji)KyrgyzLaoLatinLatvianLithuanianLuxembourgishMacedonianMalagasyMalayMalayalamMalteseMaoriMarathiMongolianMyanmar (Burmese)NepaliNorwegianPashtoPersianPolishPortuguesePunjabiRomanianRussianSamoanScottish GaelicSerbianSesothoShonaSindhiSinhalaSlovakSlovenianSomaliSpanishSundaneseSwahiliSwedishTajikTamilTeluguThaiTurkishUkrainianUrduUzbekVietnameseWelshXhosaYiddishYorubaZulu

Large Language Models are Biased Because They Are Large Language Models Philip Resnik

Large Language Models are Biased Because They Are Large Language Models

Philip Resnik

https://arxiv.org/abs/2406.13138

Obermeyer22,. Analysis of 6.4 million SARS-CoV-2 genomes identifies mutations associated with fitness.

The study "Analysis of 6.4 million SARS-CoV-2 genomes identifies mutations associated with fitness" outlines several limitations in its research:

Observational Data Limitations: The study acknowledges that while convergent evolution helps identify candidate functional mutations, observational data alone is insufficient to declare mutations as causal rather than merely correlated. Functional follow-up studies are necessary to validate the findings .
Lack of Epistatic Interactions: The PyR0 model used in the study does not incorporate epistatic interactions between mutations. Although the linear-additive model provides a foundation for inferring genetic determinants and lineage fitness, future research should include more complex modeling to account for epistatic effects and migration across geographic regions .
Regional Heterogeneity: The model showed some limitations in accounting for regional heterogeneity. Although the model's estimates were qualitatively consistent across different geographic regions, there were challenges in fully capturing the spatial structure and functional hotspots .
Bias in Lineage Assignment Process: The study notes that there may be an upward bias in fitness estimates caused by the lineage assignment process. This bias can be observed in simulation studies, which suggests that fitness values might be overestimated due to the methodology used in lineage classification .
Predictive Ability Limitations: While the model was able to forecast the rise of variants of concern reliably for 1 to 2 months into the future, its predictive ability was disrupted by the emergence of completely new strains. The forecasts stabilized typically within two weeks after the emergence of a new competitive lineage in a region .
Computational Challenges: Fitting the large model was computationally challenging, requiring the use of stochastic variational inference and a significant amount of computational resources. This complexity might limit the accessibility and reproducibility of the methodology for other researchers .

These limitations highlight areas for future improvement and the need for ongoing validation and refinement of the model to enhance its accuracy and applicability in tracking and predicting SARS-CoV-2 evolution.

https://chatgpt.com/share/c910da4a-a51f-40b1-95e6-e2a997d7af3b

Deep Learning and Classic method in GWAS

From perplexity.ai

Deep learning approaches to genome-wide association studies (GWAS) offer several advantages and disadvantages compared to classic GWAS methods. Here's a comparison of deep GWAS and classic GWAS:

Pros of deep GWAS:

1. Improved detection of complex genetic associations: Deep learning models can capture non-linear relationships and interactions between genetic variants, potentially uncovering associations that classic GWAS might miss[4].

2. Reduced multiple testing burden: Methods like DeepWAS prioritize SNPs based on functional roles, reducing the number of tests performed and potentially increasing statistical power[4].

3. Integration of functional genomic data: Deep learning approaches can incorporate additional biological information, such as regulatory data or cell-type-specific effects, to improve the interpretation of results[4].

4. Ability to handle high-dimensional data: Deep learning models are well-suited for analyzing large-scale genomic datasets with millions of variants[1].

5. Potential for improved prediction accuracy: Some deep learning methods have shown promise in improving the prediction of complex traits compared to traditional approaches[4].

Cons of deep GWAS:

1. Increased computational complexity: Deep learning models often require significant computational resources and time to train and evaluate, especially when dealing with large genomic datasets[3][4].

2. Risk of overfitting: Deep learning models with many parameters may be prone to overfitting, particularly when sample sizes are limited[1].

3. Reduced interpretability: The complex nature of deep learning models can make it challenging to interpret the biological significance of identified associations[1].

4. Dependence on training data quality: The performance of deep learning models can be sensitive to the quality and representativeness of the training data, as seen with methods like GWANN[4].

5. Potential for missing important associations: Some deep learning approaches that prioritize specific SNPs or functional units may overlook other biologically relevant associations[4].

Pros of classic GWAS:

1. Well-established methodology: Classic GWAS has a long history of use and validation in identifying genetic associations for complex traits[1][2].

2. Straightforward interpretation: Results from classic GWAS are generally easier to interpret and communicate to non-experts[1].

3. Unbiased genome-wide scanning: Classic GWAS examines associations across the entire genome without prior assumptions, potentially uncovering unexpected relationships[2].

4. Robust statistical framework: Classic GWAS relies on well-understood statistical methods for hypothesis testing and multiple comparison correction[3].

Cons of classic GWAS:

1. Limited ability to detect complex interactions: Traditional GWAS may miss non-linear relationships or epistatic interactions between genetic variants[1][4].

2. Multiple testing burden: The large number of statistical tests performed in classic GWAS requires stringent significance thresholds, potentially missing true associations with smaller effect sizes[3][4].

3. Difficulty in identifying causal variants: Classic GWAS often identifies tag SNPs in linkage disequilibrium with causal variants, requiring additional follow-up studies to pinpoint functional variants[1].

4. Limited incorporation of functional genomic information: Classic GWAS typically does not directly integrate additional biological data in the primary analysis[4].

In conclusion, while deep learning approaches to GWAS offer promising advantages in detecting complex genetic associations and integrating functional genomic data, they also come with increased computational complexity and potential interpretability challenges. Classic GWAS remains a valuable and well-established method for identifying genetic associations, but may miss some complex relationships that deep learning approaches can uncover. The choice between deep GWAS and classic GWAS depends on the specific research question, available data, and computational resources.

Citations:

[1] https://www.nature.com/articles/s43586-021-00056-9

[2] https://gwern.net/doc/genetics/heritable/2019-tam.pdf

[3] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2801749/

[4] https://www.mdpi.com/2073-4425/15/1/34

[5] https://www.youtube.com/watch?v=G2d5qaJ_g3w

responsive AI

Smits, Van der Zandt, Van Turnhous 2024

Technical: dataserts, algorithms

Interactional

Contextual: legal frameworks