Learning Flat Latent Manifolds with VAEs

Authors: Nutan Chen, Alexej Klushyn, Francesco Ferroni, Justin Bayer, Patrick Van Der Smagt

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our method on a range of data-sets, including a videotracking benchmark, where the performance of our unsupervised approach nears that of state-of-the-art supervised approaches, while retaining the computational effciency of straight-line-based ap proaches.
Researcher Affiliation Industry 1Machine Learning Research Lab, Volkswagen Group, Munich, Germany 2Autonomous Intelligent Driving Gmb H, Munich, Germany. Correspondence to: Nutan Chen <nu tan.chen@gmail.com>.
Pseudocode Yes The optimisation algorithm Alg. 1, and further details about the optimisation process can be found in App. A.4.
Open Source Code No The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper.
Open Datasets Yes We test our method on artifcial pendulum images, human motion data, MNIST, and MOT16. ... The pendulum data-set (Klushyn et al., 2019; Chen et al., 2018a) consists of 16 x 16-pixel images generated by a pen dulum simulator. ... CMU human motion data set (http://mocap.cs.cmu.edu). ... The binarised MNIST data-set (Larochelle & Murray, 2011) consists of 50,000 training and 10,000 test images... ... MOT16 object-tracking database (Milan et al., 2016)...
Dataset Splits No The binarised MNIST data-set (Larochelle & Murray, 2011) consists of 50,000 training and 10,000 test images of hand written digits (zero to nine) with 28 x 28 pixels in size.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment.
Experiment Setup Yes To be in line with previous literature (e.g. Higgins et al., 2017; Sønderby et al., 2016), we use the β-parametrisation of the Lagrange multiplier β = 1/λ in our experiments. ... VHP-FMVAE-SORT η = 300 ... VHP-FMVAE-SORT η = 3000