Thrilled that "Self-Supervised Hyperboloid Representations from Logical Queries over Knowledge Graphs" with Nikhil Rao Sumeet Karthik Subbian Chandan Reddy got accepted at The Web Conference. #TheWebConf2021 #www2021 Preprint: arxiv.org/abs/2012.13023

Today The Web Conference Ph.D. student Nurendra Choudhary Virginia Tech Computer Science The Sanghani Center at Virginia Tech presented "Self-Supervised Hyperboloid Representations Logical Queries over Knowledge Graphs," his work w/Nikhil Rao Sumeet & Karthik Subbian Amazon Science and Chandan Reddy. bit.ly/3asWkSr

My favourite irony of AI is that playing chess turns out to be fairly easy ... It's kicking a football or having a chat with your neighbour that turns out to be very hard.

Ph.D. student Nurendra Choudhary was drawn to The Sanghani Center at Virginia Tech because his research interests aligned with now advisor Chandan Reddy. He says the center has helped him broaden his work on deep learning methods in information retrieval. Read his student spotlight: bit.ly/3vsewDI

At The Web Conference, Amazon scientists presented a new way to embed knowledge graphs using hyperboloids — curved analogues of rectangles — that enables a 7% to 33% improvement over its best-performing predecessors in handling graph queries. #TheWebConf #WWW2021 amzn.to/3h9RKww

Graph neural networks are driving lots of progress in machine learning by extending deep learning approaches to complex graph data and applications. Let’s take a look at a few methods ↓

Excited for our work on Gaussian KG embeddings "Probabilistic Entity Representation Model for Chain Reasoning over Knowledge Graphs" which got accepted at #NeurIPS2021 Nikhil Rao Sumeet Karthik Subbian Chandan Reddy arxiv.org/abs/2110.13522

Excited to announce our model ANTHEM at #WSDM22 We propose a new method that utilizes hyperbolic space to answer natural language queries over the product catalogue. Hope you like it!! Nikhil Rao Sumeet Karthik Subbian @chandankreddy@AmazonScience amazon.science/publications/a…

Need a scalable approach to improve your search results. How about integrating your knowledge graphs with the queries? Take a look at our paper on Graph-based LM at #KDD2022. Paper: bit.ly/graphlm-kdd2022 Code: bit.ly/graphlm-code Amazon Science The Sanghani Center at Virginia Tech SIGKDD 2025

Excited about hyperbolic networks and want a crash course to them? Attend our tutorial Hyperbolic Neural Networks: Theory, Architectures and Applications SIGKDD 2025 #KDD2022 in Room 209C at 1 pm EDT today. Amazon Science The Sanghani Center at Virginia Tech

Presenting our work “Graph-based Multilingual Language Model: Leveraging Product Relations for Search Relevance” in room 202B at 1.30PM today at #KDD2022. Please join us if you are interested in graph-based search. Amazon Science The Sanghani Center at Virginia Tech.

Ever wanted a Gaussian process whose domain is a manifold important enough to actually have a name? If it's a Lie group or homogeneous space, we've worked out a general recipe for computing sq. exp. (heat, diffusion, ..) and Matérn kernels on it! arxiv.org/abs/2208.14960

Schrödinger's Plates: They are both broken and not broken until you open the door.

Wonder where our students The Sanghani Center at Virginia Tech go after earning a Ph.D. Virginia Tech Computer Science? Dr. Nurendra Choudhary (Nurendra Choudhary), here with advisor Chandan Reddy @VTmetroDC commencement, tells us he is joining Amazon in Palo Alto, CA, as an applied scientist II. Congrats! 🎉 #HokieGrad #VT23

It is a real pain to be constrained by the 2K context window in LLaMA models. In our arXiv, we extend the context up to 32K with <1000 fine-tuning steps and largely keep performance, by *interpolating* positional encoding, rather than extrapolating it. We also give a theoretical