#graphdatascience — Public Fediverse posts

This month makes it 2 years since I was featured as a guest on a @neo4j YouTube livestream alongside @alexandererdl, where I discussed & demoed my analysis of the FIFA22 dataset. You can watch it here, in case you didn't see the live event 2 years ago 😉

#neo4j #FIFA22 #dataviz #graphdatabase #graphDataScience

https://www.youtube.com/live/ZADwMoBJ6GQ?si=VG_HnFCvZRdYGH1h

Lately I’ve become intrigued about the published research + open source code for a relatively specific topic: generating graphs to use for inference.
Here is a comparison of five research projects circa 2019-2024 which explore different ways of generating graphs to use for inference.

https://blog.derwen.ai/graphs-for-inference-684d73d8b59c

#graphDataScience

Lately I’ve become intrigued about the published research + open source code for a relatively specific topic: generating graphs to use for inference.
Here is a comparison of five research projects circa 2019-2024 which explore different ways of generating graphs to use for inference.

https://blog.derwen.ai/graphs-for-inference-684d73d8b59c

#graphDataScience

Visualizing a document vector embedding (index) as a clustered k-nearest-neighbour graph is so insightful. You see which of those 20k+ arxiv papers are close together by the text embedding of their abstract. #neo4j #graphdatascience and you can expand from the vector search to the context of your documents (authors, venues, categories and related information) to e.g. power a #rag application.

Visualizing a document vector embedding (index) as a clustered k-nearest-neighbour graph is so insightful. You see which of those 20k+ arxiv papers are close together by the text embedding of their abstract. #neo4j #graphdatascience and you can expand from the vector search to the context of your documents (authors, venues, categories and related information) to e.g. power a #rag application.

"Graph Levels of Detail"
https://blog.derwen.ai/graph-levels-of-detail-ea4226abba55

We're circulating for review this survey of methods for abstraction layers in knowledge graphs. This covers mathematical approaches from several areas, so it's important to hear back whether any of the descriptions are misrepresented. Also, is this kind of work interesting for your organization?

#graphDataScience

"Graph Levels of Detail"
https://blog.derwen.ai/graph-levels-of-detail-ea4226abba55

We're circulating for review this survey of methods for abstraction layers in knowledge graphs. This covers mathematical approaches from several areas, so it's important to hear back whether any of the descriptions are misrepresented. Also, is this kind of work interesting for your organization?

#graphDataScience

Here are slides for my recent talk -- I really appreciated the opportunity to present at K1st World, JK2K's meetup in DC, and Corunna Innovation Summit, and the many interesting discussions!

"Language, Graphs, and AI in Industry"
https://derwen.ai/s/mqqm

This links to a directory of resources related to graph resources:
https://derwen.ai/graph/

Also, join our Graph Data Science group on LinkedIn:
https://www.linkedin.com/groups/6725785/

#graphDataScience

Here are slides for my recent talk -- I really appreciated the opportunity to present at K1st World, JK2K's meetup in DC, and Corunna Innovation Summit, and the many interesting discussions!

"Language, Graphs, and AI in Industry"
https://derwen.ai/s/mqqm

This links to a directory of resources related to graph resources:
https://derwen.ai/graph/

Also, join our Graph Data Science group on LinkedIn:
https://www.linkedin.com/groups/6725785/

#graphDataScience

A summary of links for the emerging intersection of graph ML and uses for LLMs

https://blog.derwen.ai/visual-missives-from-the-latent-space-2023-10-16-d4bfa944b86c

#graphDataScience

A summary of links for the emerging intersection of graph ML and uses for LLMs

https://blog.derwen.ai/visual-missives-from-the-latent-space-2023-10-16-d4bfa944b86c

#graphDataScience

New 3.2.5 release of `pytextrank` is available on PyPi.org now. Kudos to @hedgeknight @normanbaatz
https://pypi.org/project/pytextrank/3.2.5/
#python #nlp #opensource #graphDataScience

New 3.2.5 release of `pytextrank` is available on PyPi.org now. Kudos to @hedgeknight @normanbaatz
https://pypi.org/project/pytextrank/3.2.5/
#python #nlp #opensource #graphDataScience

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs

#nlproc #machinelearning #generativeai
🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs
#nlproc #machinelearning #generativeai

🔗 https://thedataexchange.media/the-future-of-graph-databases/

#TheDataExchangePod 🎧 Emil Eifrem of Neo4j unlocks the secrets of #GraphDatabases, #LLMs, #VectorDatabases, and more. We examine the interplay of these elements with knowledge graphs and applications of graph neural networks. We explore the rise of new database companies and delve into the world of #GraphDataScience & retrieval-augmented LLMs
#nlproc #machinelearning #generativeai

🔗 https://thedataexchange.media/the-future-of-graph-databases/

"GraphPrompt: Unifying Pre-Training and Downstream Tasks for Graph Neural Networks"
https://arxiv.org/abs/2302.08043
Zemin Liu, et al. (2023-02-16)

A novel pre-training and prompting framework on graphs, for the "pre-train, fine-tune" and "pre-train, prompt" approach to working with GNNs

From the excellent "Prompt Engineering Guide" by Elvis Savaria
https://github.com/dair-ai/Prompt-Engineering-Guide

#graphDataScience

"GraphPrompt: Unifying Pre-Training and Downstream Tasks for Graph Neural Networks"
https://arxiv.org/abs/2302.08043
Zemin Liu, et al. (2023-02-16)

A novel pre-training and prompting framework on graphs, for the "pre-train, fine-tune" and "pre-train, prompt" approach to working with GNNs

From the excellent "Prompt Engineering Guide" by Elvis Savaria
https://github.com/dair-ai/Prompt-Engineering-Guide

#graphDataScience

exploring diffusion in GraphML, in drug discovery:

"Denoising Diffusion Generative Models in Graph ML"
https://towardsdatascience.com/denoising-diffusion-generative-models-in-graph-ml-c496af5811c5
Michael Galkin
Towards Data Science (2022-11-26)

#graphDataScience

exploring diffusion in GraphML, in drug discovery:

"Denoising Diffusion Generative Models in Graph ML"
https://towardsdatascience.com/denoising-diffusion-generative-models-in-graph-ml-c496af5811c5
Michael Galkin
Towards Data Science (2022-11-26)

#graphDataScience

a very good introduction to GraphML:

"Introduction to Graph Machine Learning"
https://huggingface.co/blog/intro-graphml
Clémentine Fourrier, Hugging Face

#graphDataScience

a very good introduction to GraphML:

"Introduction to Graph Machine Learning"
https://huggingface.co/blog/intro-graphml
Clémentine Fourrier, Hugging Face

#graphDataScience

excellent visualizations about graph algorithms:
"The hidden beauty of the A* algorithm"
https://www.youtube.com/watch?v=A60q6dcoCjw

from Luis Natera's newsletter https://buttondown.email/natera

#graphDataScience

excellent visualizations about graph algorithms:
"The hidden beauty of the A* algorithm"
https://www.youtube.com/watch?v=A60q6dcoCjw

from Luis Natera's newsletter https://buttondown.email/natera

#graphDataScience

Year-end analysis/trend estimates by leading analysts/doers/thinkers -

Ben Lorica, Mikio Braun, Jenn Webb
The Data Exchange
https://thedataexchange.media/2023-opportunities-and-trends-data-machine-learning-and-ai/

> use of generative AI and language models have become dominant trends.

Jeff Dean
Google Research
https://ai.googleblog.com/2023/01/google-research-2022-beyond-language.html

> advances in generative AI, multimodel models, stable diffusion

#graphdatascience

Year-end analysis/trend estimates by leading analysts/doers/thinkers -

Ben Lorica, Mikio Braun, Jenn Webb
The Data Exchange
https://thedataexchange.media/2023-opportunities-and-trends-data-machine-learning-and-ai/

> use of generative AI and language models have become dominant trends.

Jeff Dean
Google Research
https://ai.googleblog.com/2023/01/google-research-2022-beyond-language.html

> advances in generative AI, multimodel models, stable diffusion

#graphdatascience

Many of the keynote and session videos for PyData Global 2022 went up online today, and here's my talk:
https://youtu.be/IKFGFFtxgow?t=5463

#graphdatascience

Many of the keynote and session videos for PyData Global 2022 went up online today, and here's my talk:
https://youtu.be/IKFGFFtxgow?t=5463

#graphdatascience

"Universality of Neural Networks on Graphs vs. Sets"
Petar Veličković, Fabian Fuchs
2022-11
https://fabianfuchsml.github.io/universalgraphs/

Looking at universal function approximation (in deep learning) applied to graphs.

More so about universal function *representation* than *approximation*

Also, what is provably non-universal? For example, GCNs.

#graphthinking #graphdatascience

"Universality of Neural Networks on Graphs vs. Sets"
Petar Veličković, Fabian Fuchs
2022-11
https://fabianfuchsml.github.io/universalgraphs/

Looking at universal function approximation (in deep learning) applied to graphs.

More so about universal function *representation* than *approximation*

Also, what is provably non-universal? For example, GCNs.

#graphthinking #graphdatascience

"From Knowledge Graphs to Knowledge Categories"
Josh Shinavier interviews Ryan Wisnesky
https://www.youtube.com/watch?v=-N33MZa3B9o

Applications of category theory with graphs. For example, how to align schema, make guarantees about data migration from relational databases into graphs, data quality checks, etc. If you've ever worked in some of these areas of advanced math, Ryan shows excellent applications – including some of the data management practices at Uber.

#graphthinking #graphdatascience

"From Knowledge Graphs to Knowledge Categories"
Josh Shinavier interviews Ryan Wisnesky
https://www.youtube.com/watch?v=-N33MZa3B9o

Applications of category theory with graphs. For example, how to align schema, make guarantees about data migration from relational databases into graphs, data quality checks, etc. If you've ever worked in some of these areas of advanced math, Ryan shows excellent applications – including some of the data management practices at Uber.

#graphthinking #graphdatascience

I'll present at PyData Global, Thu Dec 01 13:30 US Pacific:
"Data Prep for Graphs"
https://global2022.pydata.org/cfp/talk/AH9DJD/

TL;DR: data prep phase in #graphdatascience work involves tools/techniques vastly different than data science in general. This stage of work is computationally expensive, and ironically much must be performed *prior* to loading into a graph DB.

Here's a sampler.

Also, we'll cover the https://github.com/DerwenAI/pynock proposal for Parquet serialization of graph data.

#graphthinking

I'll present at PyData Global, Thu Dec 01 13:30 US Pacific:
"Data Prep for Graphs"
https://global2022.pydata.org/cfp/talk/AH9DJD/

TL;DR: data prep phase in #graphdatascience work involves tools/techniques vastly different than data science in general. This stage of work is computationally expensive, and ironically much must be performed *prior* to loading into a graph DB.

Here's a sampler.

Also, we'll cover the https://github.com/DerwenAI/pynock proposal for Parquet serialization of graph data.

#graphthinking

AWS updates Neptune to provide features for real-time inductive inference of Graph Neural Networks

https://www.linkedin.com/posts/denisekgosnell_amazon-neptune-engine-version-1202-2022-activity-7003086073417342977-oPsg

https://docs.aws.amazon.com/neptune/latest/userguide/engine-releases-1.2.0.2.html

#graphthinking #graphdatascience

AWS updates Neptune to provide features for real-time inductive inference of Graph Neural Networks

https://www.linkedin.com/posts/denisekgosnell_amazon-neptune-engine-version-1202-2022-activity-7003086073417342977-oPsg

https://docs.aws.amazon.com/neptune/latest/userguide/engine-releases-1.2.0.2.html

#graphthinking #graphdatascience

KGC 2023 was recently announced.
This conference will be held May 8-12 in NYC
https://derwen.ai/s/7cypd7vy96hk

The CFP is open at:
https://docs.google.com/forms/d/1Qd3bUHAGI9JFcsuxLvIpsWYQjr5XPm_I15Y1xzneaJ4/viewform?edit_requested=true

Highly recommended.

#graphthinking #graphdatascience

KGC 2023 was recently announced.
This conference will be held May 8-12 in NYC
https://derwen.ai/s/7cypd7vy96hk

The CFP is open at:
https://docs.google.com/forms/d/1Qd3bUHAGI9JFcsuxLvIpsWYQjr5XPm_I15Y1xzneaJ4/viewform?edit_requested=true

Highly recommended.

#graphthinking #graphdatascience

A synthetic taxonomy for classifying the plastic tags from bread and other plastic-bagged pastries.

https://www.inverse.com/input/culture/horg-plastic-bread-tags-occlupanids-classification-site

> “It really STRUCK me how weirdly biomorphic it looks, like a larval PARASITE with claws. Why does no one NOTICE these things?”

#graphthinking #graphdatascience

A synthetic taxonomy for classifying the plastic tags from bread and other plastic-bagged pastries.

https://www.inverse.com/input/culture/horg-plastic-bread-tags-occlupanids-classification-site

> “It really STRUCK me how weirdly biomorphic it looks, like a larval PARASITE with claws. Why does no one NOTICE these things?”

#graphthinking #graphdatascience

Definitely, check out the amazing work by Yalda Shankar at the nexus of AI and Design:
https://yaldashankar.org/
https://www.linkedin.com/feed/update/urn:li:activity:7001187370481893376/

In particular, see "The GNN Booklet" (part 1, WIP) for an outstanding illustrated review of graph-related concepts and the associated math:
https://yaldashankar.org/index.html#Writings

#graphthinking #graphdatascience

Definitely, check out the amazing work by Yalda Shankar at the nexus of AI and Design:
https://yaldashankar.org/
https://www.linkedin.com/feed/update/urn:li:activity:7001187370481893376/

In particular, see "The GNN Booklet" (part 1, WIP) for an outstanding illustrated review of graph-related concepts and the associated math:
https://yaldashankar.org/index.html#Writings

#graphthinking #graphdatascience

https://youtu.be/cBNDpmAm78U

Using graph algorithms for module detection in open source software

#neo4j
#graphdatascience
#opensourcesecurity
#socialcyber

https://youtu.be/cBNDpmAm78U

Using graph algorithms for module detection in open source software

#neo4j
#graphdatascience
#opensourcesecurity
#socialcyber