Scalable Adaptive Stochastic Optimization Using Random Projections

Authors: Gabriel Krummenacher, Brian McWilliams, Yannic Kilcher, Joachim M. Buhmann, Nicolai Meinshausen

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 4 Experiments, We compare the performance of our proposed algorithms against both the diagonal and full-matrix ADAGRAD variants in the idealised setting where the data is dense but has low effective rank. Figure 2: Comparison of training loss (top row) and test accuracy (bottom row) on (a) MNIST, (b) CIFAR and (c) SVHN.
Researcher Affiliation Collaboration Institute for Machine Learning, Department of Computer Science, ETH Zürich, Switzerland Seminar for Statistics, Department of Mathematics, ETH Zürich, Switzerland Disney Research, Zürich, Switzerland
Pseudocode Yes Algorithm 1 ADA-LR, Algorithm 2 RADAGRAD
Open Source Code No The paper does not provide explicit statements or links indicating that the source code for the described methodology is publicly available.
Open Datasets Yes MNIST, CIFAR-10 and SVHN datasets. We trained and evaluated our network on the Penn Treebank dataset [25].
Dataset Splits No For each algorithm learning rates are tuned using cross validation. Step sizes were determined by coarsely searching a log scale of possible values and evaluating performance on a validation set. (Explanation: While validation and cross-validation are mentioned, specific split percentages, sample counts, or explicit methodologies for creating these splits are not provided in the text.)
Hardware Specification No The paper mentions general use of GPUs but does not provide specific hardware details such as GPU/CPU models, memory specifications, or detailed computer configurations used for the experiments.
Software Dependencies No The paper mentions using the 'FFTW package' but does not specify its version number or any other software dependencies with explicit version information.
Experiment Setup Yes We used a batch size of 8 and trained the networks without momentum or weight decay, in order to eliminate confounding factors. Instead, we used dropout regularization (p = 0.5) in the dense layers during training. Step sizes were determined by coarsely searching a log scale of possible values and evaluating performance on a validation set. The memory size of the T-LSTM units was set to 256.