SpectralNet: Spectral Clustering using Deep Neural Networks

Authors: Uri Shaham, Kelly Stanton, Henry Li, Ronen Basri, Boaz Nadler, Yuval Kluger

ICLR 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments indicate that our network indeed approximates the Laplacian eigenvectors well, allowing the network to cluster challenging non-convex point sets, which recent deep network based methods fail to handle; see examples in Figure 1. Finally, Spetral Net achieves competitive performance on MNIST handwritten digit dataset and state-of-the-art on the Reuters document dataset, whose size makes standard spectral clustering inapplicable.
Researcher Affiliation Academia Uri Shaham , Kelly Stanton , Henry Li Yale University New Haven, CT, USA {uri.shaham, kelly.stanton, henry.li}@yale.edu Boaz Nadler, Ronen Basri Weizmann Institute of Science Rehovot, Israel {boaz.nadler, ronen.basri}@gmail.com Yuval Kluger Yale University New Haven, CT, USA yuval.kluger@yale.edu
Pseudocode Yes Algorithm 1: Spectral Net training
Open Source Code Yes Our implementation is publicly available at https://github.com/kstant0725/Spectral Net.
Open Datasets Yes Our experiments indicate that our network indeed approximates the Laplacian eigenvectors well, allowing the network to cluster challenging non-convex point sets, which recent deep network based methods fail to handle; see examples in Figure 1. Finally, Spetral Net achieves competitive performance on MNIST handwritten digit dataset and state-of-the-art on the Reuters document dataset, whose size makes standard spectral clustering inapplicable.
Dataset Splits Yes The learning rate policy for all nets was determined by monitoring the loss on a validation set (a random subset of the training set); once the validation loss did not improve for a specified number of epochs (see patience epochs in Table 3), we divided the learning rate by 10 (see LR decay in Table 3).
Hardware Specification Yes Our Spectral Net implementation took less than 20 minutes to learn the spectral map on this dataset, using a Ge Force GTX 1080 GPU.
Software Dependencies No The paper mentions 'Python s sklearn.cluster' and 'ARPACK' but does not provide specific version numbers for these or any other software components.
Experiment Setup Yes The architectures of the Siamese net and Spectral Net are described in Table 2. Additional technical details are shown in Table 3.