Unveiling the Cycloid Trajectory of EM Iterations in Mixed Linear Regression
Authors: Zhankun Luo, Abolfazl Hashemi
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section of empirical experiments, we validate the theoretical findings established in the preceding sections. From a normal distribution N(0, Id), we sample 5,000 independent and identically distributed (i.i.d.) d-dimensional covariates, denoted as {xi}n i=1. The true parameters θ are randomly chosen from a d-dimensional unit sphere. |
| Researcher Affiliation | Academia | 1School of Electrical and Computer Engineering, Purdue University, IN, USA. |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks (e.g., sections explicitly labeled "Pseudocode" or "Algorithm"). |
| Open Source Code | Yes | The code for numerical experiments is available at https://github.com/dassein/cycloid_em_mlr. |
| Open Datasets | No | The paper states, "From a normal distribution N(0, Id), we sample 5,000 independent and identically distributed (i.i.d.) d-dimensional covariates, denoted as {xi}n i=1." It describes a data generation process rather than using a publicly available or open dataset, and does not provide access information for the generated data. |
| Dataset Splits | No | The paper states, "In all experiments, we utilize the entire dataset for EM updates at every iteration." This indicates that specific training, validation, and test splits are not explicitly used in the traditional sense; rather, the full dataset is processed in each iteration. |
| Hardware Specification | No | The paper does not provide specific hardware details such as CPU, GPU models, memory, or cloud computing instance types used for running the experiments. It only mentions general experimental procedures. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., Python, PyTorch, or other libraries with their particular versions) that would be needed to replicate the experimental environment. |
| Experiment Setup | No | The paper describes generating data, uniformly initializing parameters and mixing weights, and averaging results over 50 trials: "We uniformly choose the initial values for the parameters and the mixing weights from a unit sphere and the interval [0, 1], respectively. All the points of 4 EM iterations in Fig. 4a are the average of 50 trials with different initial values." However, it does not provide concrete details for the experimental setup such as specific hyperparameters (e.g., learning rate, batch size) or optimizer settings for the EM algorithm itself. |