Ultrahyperbolic Representation Learning

Authors: Marc Law, Jos Stam

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our novel framework is applied to graph representations. We now experimentally validate our proposed optimization methods and the effectiveness of our dissimilarity function.
Researcher Affiliation Industry This article was entirely funded by NVIDIA corporation. Marc Law and Jos Stam completed this working from home during the COVID-19 pandemic.
Pseudocode Yes Algorithm 1 Pseudo-Riemannian optimization on Qp,q β
Open Source Code No The paper does not provide any concrete access information (e.g., repository link, explicit statement of code release) for the methodology's source code.
Open Datasets Yes Zachary s karate club dataset [30]. Due to lack of space, we also report in the supp. material similar experiments on a larger hierarchical dataset [9] that describes co-authorship from papers published at NIPS from 1988 to 2003.
Dataset Splits No The paper does not specify exact percentages or sample counts for training, validation, or test splits, nor does it refer to predefined splits from cited works.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No We coded our approach in Py Torch [22] that automatically calculates the Euclidean gradient f(xi). While PyTorch is mentioned, a specific version number is not provided.
Experiment Setup Yes Initially, a random set of vectors {zi}n i=1 is generated close to the positive pole ( p |β|, 0, , 0) Qp,q β with every coordinate perturbed uniformly with a random value in the interval [ ε, ε] where ε > 0 is chosen small enough so that zi 2 q < 0. We set β = 1, ε = 0.1 and τ = 10 2. Initial embeddings are generated as follows: i, xi = p | zi 2 q| Qp,q β . In each test, we vary the number of time dimensions q + 1 while the ambient space is of fixed dimensionality d = p + q + 1 = 10.