Reduce LLM hallucinations with RAG over textual as well as structured knowledge bases. Together with Amazon we are releasing 🌟STaRK 🌟, a large-scale LLM retrieval benchmark on semi-structured knowledge bases with dataset from e-commerce, biomedicine, and academic research.

In a patient explainer, Giuseppe Futia proposes potential solutions for integrating LLMs, knowledge graphs, and ontologies to enable a virtuous cycle to consistently improve named-entity disambiguation systems in the biomedical domain. buff.ly/3UuiH1p

Graph Machine Learning in the Era of LLMs

Provides a comprehensive overview of the latest advancements for Graph ML in the era of LLMs.

Covers the recent developments in Graph ML, how LLM can enhance graph features, and how it can address issues such as OOD and graph…

Check out my new article for Towards Data Science entitled:

'Leveraging #AI Synergies for Named Entity Disambiguation'.

The article describes how to harness the synergy between #LLMs , #KnowledgeGraphs , and #ontologies to pave the way for enhanced NED systems in biomedicine.

We started releasing the first chapters of our Geometric Deep Learning book and the accompanying slides from the corresponding Oxford and Cambridge courses.

Notebooks for Lecture 4 on Graph SVM

Lab1: Standard/Linear SVM
github.com/xbresson/GML20…

Lab2: Soft-Margin SVM
github.com/xbresson/GML20…

Lab3: Kernel/Non-Linear SVM
github.com/xbresson/GML20…

Lab4: Graph SVM
github.com/xbresson/GML20…

Graph Machine Learning course

Lecture 4 presents Graph SVM 🌟

The celebrated Support Vector Machine (SVM) coupled with kernel method and graph Dirichlet regularization is one of the best classification models (in the absence of feature learning :)

storage.googleapis.com/xavierbresson/…

What is Neurosymbolic AI, and how are Knowledge Graphs a part of it?

Knowledge Graphs are by far the largest scale knowledge representation formalism in the history of AI. Graphs of hundreds of millions of triples are now routinely used in both academic research and industrial…

🔍 How to generalize queries on any arbitrary KG in a zero-shot fashion?

This is the purpose of the paper released by Michael Galkin, Jincheng Zhou, Bruno Ribeiro, Jian Tang, and Zhaocheng Zhu, who present ULTRAQUERY. arxiv.org/abs/2404.07198

GNN+TSP+Genome 🚀

We have pioneered a graph learning approach for genome reconstruction, marking a significant milestone in AI+Biology.

Great interdisciplinary team, with Lovro Vrček at the helm!

Paper, code, data have been released to empower further progress. 👇

Our GitHub repo Awesome-Graph-LLM: github.com/XiaoxinHe/Awes…, has received 1k+ stars! 🌟A huge thank you to everyone who has supported and contributed. Let's keep pushing the boundaries of Awesome-Graph-LLM together!

Register for the workshop on LLMs, GNNs and Foundation Models -- it will be exciting!
The Web Conference #WebConference2024 👇

gRNAde is now on Colaboratory 🙌

You can do fixed backbone re-design of your own RNA 3D structures from PDB files right from your browser! (ProteinMPNN for RNA structure)

📝gRNAde 101 tutorial: colab.research.google.com/drive/16rXKgbG…

🚀Design notebook for new tasks: colab.research.google.com/drive/1ajcikLb…

Notebooks for Lecture 2 on Graph Science

Lab1: Generate LFR social networks
github.com/xbresson/GML20…

Lab2: Visualize spectrum of point cloud & grid
github.com/xbresson/GML20…

Lab3/4: Graph construction for two-moon & text documents
github.com/xbresson/GML20…
github.com/xbresson/GML20…

🚀 GPTFast: Accelerate your LLMs 6-7x

PyTorch released an excellent blog post (pytorch.org/blog/accelerat…) in November last year about how to accelerate LLMs using different techniques:

- Torch.compile: Compiler for PyTorch models
- GPU quantization: Reduce weight precision
-…

'Graphs represent universal models to describe interacting elements, and Graph Neural Networks (GNNs) have become an essential toolkit for applying learning algorithms to graph-structured data.' Giuseppe Futia unveils a new series on GNNs' inner workings. buff.ly/42NiSHr

📣 Check out my new article about “The Expressive Power of GNNs” on Towards Data Science.

🔎 The first article of this series is dedicated to the foundational concepts related to the representation of graphs and the notion of graph isomorphism.

medium.com/p/5cdb4bca6ae3

New (2h13m 😅) lecture: 'Let's build the GPT Tokenizer'

Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set, training algorithm (Byte Pair Encoding), and after training implement two functions: encode() from strings to tokens, and…

Are you ready to turbocharge your neural network training? 🌟 My latest article unveils the magic of Batch Normalization, a must-know for any aspiring data scientist. #BatchNormalization #AI #DeepLearning #TechTips #DataScience
dataleadsfuture.com/visualizing-wh…

Ready to dive into the fascinating world of graph neural networks? Giuseppe Futia just launched a series exploring the theory behind GNNs as well as their practical applications. Part one lays the foundations and covers GNNs' essential building blocks. buff.ly/42NiSHr