Learning Low-dimensional Latent Dynamics from High-dimensional Observations: Non-asymptotics and Lower Bounds

Authors: Yuyang Zhang, Shahriar Talebi, Na Li

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We simulate our algorithm for a set of simple systems where r = m = 1, Σw = 0 and ση = 1. We use A = 0.9, B = 1 and randomly sample C with orthonormal columns. We choose inputs with covariance Σu = 0.1. The choice of n will be clear in the context. The results are reported in Figure 1.
Researcher Affiliation Academia Yuyang Zhang 1 Shahriar Talebi 1 Na Li 1 1SEAS, Harvard University, Cambridge, USA.
Pseudocode Yes Algorithm 1 Column Space Projection SYSID (Col-SYSID) Algorithm 2 Column Space Approximation (Col-Approx) Algorithm 3 Meta Column Space Projection SYSID (Meta-Col-SYSID) Algorithm 4 Ho-Kalman
Open Source Code No The paper does not contain any statements about making its source code publicly available, nor does it provide a link to a code repository.
Open Datasets No The paper uses simulated data for its experiments, stating, 'We simulate our algorithm for a set of simple systems where r = m = 1, Σw = 0 and ση = 1.' It does not refer to any pre-existing public datasets with access information.
Dataset Splits Yes We consider system M and datasets D1 = U1 Y1, D2 = U2 Y2 (with lengths T1, T2 respectively) in HDSYSID. ... If T1 κ3 n2r3, T2 κ1 poly (r, m)...
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory, or cluster specifications) used to run the experiments.
Software Dependencies No The paper does not list any specific software dependencies, libraries, or solvers with version numbers.
Experiment Setup Yes We simulate our algorithm for a set of simple systems where r = m = 1, Σw = 0 and ση = 1. We use A = 0.9, B = 1 and randomly sample C with orthonormal columns. We choose inputs with covariance Σu = 0.1. ... We first simulate Col-Approx (Algorithm 2) with n = 40, 80, 160, 320 and a single trajectory data of length T = 10000 separately (Figure 1.Left). Next, we simulate Col-SYSID (Algorithm 1) with T = 10000 and n = 320 (Figure 1.Center).