Deep Homogeneous Mixture Models: Representation, Separation, and Approximation

Authors: Priyank Jaini, Pascal Poupart, Yaoliang Yu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on both synthetic and real datasets confirm the benefits of depth in density estimation.
Researcher Affiliation Academia Priyank Jaini Department of Computer Science & Waterloo AI Institute University of Waterloo pjaini@uwaterloo.ca Pascal Poupart University of Waterloo, Vector Institute & Waterloo AI Institute ppoupart@uwaterloo.ca Yaoliang Yu Department of Computer Science & Waterloo AI Institute University of Waterloo yaoliang.yu@uwaterloo.ca
Pseudocode No The paper describes algorithms but does not provide pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper mentions adapting code from another work ('HT-TMM') but does not state that its own developed code ('SPN-CG') is open-source or provide a link to it.
Open Datasets Yes We perform experiments on MNIST [15] for digit classification and small NORB [16] for 3D object recognition
Dataset Splits No The paper mentions using MNIST and NORB datasets and adapting experiments from [26], but it does not explicitly state the training, validation, or test splits used within its text.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments.
Software Dependencies No The paper mentions using an 'Adam SGD variant' for training but does not specify any software packages, libraries, or their version numbers.
Experiment Setup Yes For each iteration, we train the network using an Adam SGD variant with a base learning rate of 0.03 and momentum parameters β1 = β2 = 0.9. For each added network structure, we train the model for 22,000 iterations for MNIST and 40,000 for NORB.