Provable Subspace Identification Under Post-Nonlinear Mixtures

Authors: Qi Lyu, Xiao Fu

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental A series of numerical experiments corroborate our theoretical claims.
Researcher Affiliation Academia Qi Lyu School of EECS Oregon State University Corvallis, OR 97331 lyuqi@oregonstate.edu Xiao Fu School of EECS Oregon State University Corvallis, OR 97331 xiao.fu@oregonstate.edu
Pseudocode Yes Algorithm 1: Post-Nonlinear Subspace Identification.
Open Source Code Yes The source code can be found online2. https://github.com/llvqi
Open Datasets Yes We use the human face electroencephalogram (EEG) dataset3. https://www.fil.ion.ucl.ac.uk/spm/data/mmfaces/
Dataset Splits Yes We have N = 27, 692 samples of x , which are split as training, validation and test sets with 24794, 1449, and 1449 samples, respectively.
Hardware Specification No The paper does not specify the exact hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions 'Adam optimizer [30]' but does not provide specific version numbers for Adam or any other software dependencies.
Experiment Setup Yes For the neural networks representing fm( ) and rm( ), we use R = 256 with Re LU activations. We use the Adam optimizer [30] with the initial learning rate being 2e 4 for the network optimization part. For the hyperparameters, we set λ = 1e 4.