Harmonic Exponential Families on Manifolds

Authors: Taco Cohen, Max Welling

ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results show that harmonic densities yield a significantly higher likelihood than the best competing method, while being orders of magnitude faster to train.
Researcher Affiliation Academia Taco S. Cohen T.S.COHEN@UVA.NL University of Amsterdam Max Welling M.WELLING@UVA.NL University of Amsterdam University of California Irvine Canadian Institute for Advanced Research
Pseudocode Yes So we have an efficient algorithm for computing moments: 1. Compute ϕ = exp (F 1η). 2. Compute M = F ϕ 3. Compute Ep(g|η) [T(g)] = M/M 0 00. [...] To find the optimal transformation, first perform posterior inference (steps 1 and 2) and then maximize (step 3): 1. Compute ˆx = F x and ˆy = F y. 2. Compute ηλ = ηλ + 1 σ2 dim λ ˆxλˆy T λ 3. Compute g = arg maxi[F 1 η](gi)
Open Source Code No The paper does not provide concrete access to its own source code.
Open Datasets Yes We obtained the Significant Earthquake Dataset (NGDC, 2015) from the National Geophysical Datacenter of the National Oceanographic and Athmospheric Administration.
Dataset Splits Yes Figure 2 shows the average train and test log-likelihood over 5 cross-validation folds, for the spherical harmonic density and the mixture of Kent distribution.
Hardware Specification No The paper does not provide specific details about the hardware used for running its experiments.
Software Dependencies No The paper mentions software like Python, NFFT library, SciPy routines, and L-BFGS algorithm, but does not provide specific version numbers for these dependencies.
Experiment Setup No The paper describes regularization methods and optimization algorithms, but does not provide specific concrete hyperparameter values (e.g., learning rate, batch size) or detailed training configurations for its experiments.