On the Stochastic Stability of Deep Markov Models
Authors: Jan Drgona, Sayak Mukherjee, Jiaxin Zhang, Frank Liu, Mahantesh Halappanavar
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically substantiate our theoretical results via intuitive numerical experiments using the proposed stability constraints. In this section we empirically validate the conditions given in Theorem 3 and Corollary 5 by investigating the dynamics of DMM s transition maps (12) whose mean fθf (x) and variance gθg(x) are parametrized by neural networks with different spectral distributions of their weights and activation scaling matrices (6). |
| Researcher Affiliation | Academia | 1 Pacific Northwest National Laboratory Richland, Washington, USA 2 Oak Ridge National Laboratory Oak Ridge, Tennessee, USA |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The authors plan to open-source the code with the camera ready version of the paper. The reason why we do not disclose the case study code during the review process is the use of custom open-source libraries that could reveal the affiliation of the authors. |
| Open Datasets | No | The paper states that experiments were performed using Pyro, a probabilistic programming language, and describes generating DMMs with different properties for numerical case studies, but it does not specify the use of a publicly available dataset or provide concrete access information to any dataset used for training. |
| Dataset Splits | No | The paper describes numerical experiments and visualizations of DMM dynamics, but it does not provide specific details on data splits for training, validation, or testing. |
| Hardware Specification | No | The paper states in the ethics checklist: "Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [N/A]" |
| Software Dependencies | No | The paper mentions using "the probabilistic programming language Pyro [Bingham et al., 2019]", but it does not specify a version number for Pyro or any other software dependencies. |
| Experiment Setup | No | The paper discusses the design of experiments in Section 4.1, mentioning variations in activation functions, layer depth, and bias presence, and different weight initialization methods. However, it states in the ethics checklist: "Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] We provide the summary of the experiment hyperparameters in the supplementary material." indicating that specific hyperparameter values are not in the main text. |