GenAI embedding spaces - collecting papers, libraries and thoughts on genAI computation centered on embedding spaces

Journal Club

Your favorite reference for models you want the rest of us to be familiar with:

• Praful : Path genAI are you using ...
  https://www.nature.com/articles/s41591-024-02857-3  
  https://www.nature.com/articles/s41591-024-02856-4 

• Jonas : Advancing Multimodal Medical Capabilities of Gemini
  Capabilities of Gemini Models in Medicine
  https://research.google/blog/advancing-medical-ai-with-med-gemini 

• Daniel cGAN: https://arxiv.org/pdf/1411.1784 

• ...

Getting math serious about embedded spaces

Properties:

1. Embedding is a lossy encoder, at best it facilitates the definition of metric spaces.

2. Define a dissipative system, decoding not straight froward

3. Define metric spaces for multivariate exploratory statistical methods ( Lee may call it EDA, but he's in Japan now :-D). See also Cluster Radar and epiVECS.

4. Projectors ...

5. Praful modelling embeded space

6. Daniel modelling the embedded space

7. Anyone else has embedded modelling, or thoughts on how to do it, to show and tell?

8. ARPAH distributed file system ... https://epiverse.github.io/catalog/catalog/index.json 

9. https://github.com/jeyabbalas/vector-gazer <-- in-browser

10. https://www.tensorflow.org/tensorboard/tensorboard_projector_plugin <-- in-Python

11. ...

Hackathon

TensorflowProjector

https://projector.tensorflow.org, should we use it as a general purpose display platform?

Inês TCGA-Reports embedded

Jeya's telescope

... vector gazing...

https://github.com/jeyabbalas/vector-gazer 

https://jeyabbalas.github.io/vector-gazer/ 

Ines is using Plotly ...

https://epiverse.github.io/tcgapath 

Modelling

Recalling that embeddings are being used for something ... such as modelling

<silly> Embedding as an advance, non-binary, version of one hot encoding

Recall multivariate exploratory statistics and the metrics that enable them (i.e. metric spaces)

...

WebR update

https://rud.is/w/webr-pkgs

https://quarto-webr.thecoatlessprofessor.com/examples/readme

...

11:00 - Data Science Seminar Series: Dr. Renato Umeton

Wednesday, November 13, 2024 11:00 AM - 12:00 PM
(UTC-05:00) Eastern Time (US & Canada)
Registration for this webinar

Integrating AI and Generative AI in Operations and Research: From Creating the Team and Infrastructure to Supporting 2,000+ Users

In this webinar, Dana-Farber Cancer Institute’s Dr. Renato Umeton will explore the transformative journey of applying artificial intelligence (AI) solutions at scale, and in full compliance, bridging the gap between Big Tech, operations, and research.

the deployment of GPT4DFCI—a generative AI tool, restricted to the Dana-Farber private network, that's helping the workforce with operational, administrative, and research uses. (This does not include direct clinical care.)

Throughout this hour, learn about challenges faced, breakthroughs achieved, and the future directions of expanding AI capabilities in your cancer research.