Automatic dimensionality selection from the scree plot via the use of profile likelihood

By Mu Zhu, A. Ghodsi. Computational Statistics & Data Analysis, 2006

453 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/d3a556c96ed6f170823006c1b3a043f139f9bd7d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Variable-Centered, Person-Centered, and Person-Specific Approaches (2018, methodology): “First, the scree plot was visually analyzed for an “elbow” or sharp decrease in eigenvalues, which suggests that factors after the elbow provided little explanatory information compared to those prior to the elbow (Nasser, Benson, & Wisenbaker, 2002; Zhu & Ghodsi, 2006).”
Statistical Inference on Random Dot Product Graphs: a Survey (2017, methodology): “3, we discuss approaches (see [18] and [109]) to estimating the embedding dimension.”
Atlas-Based Quantification of Cardiac Remodeling Due to Myocardial Infarction (2014, unknown): “A scree plot [29] is given in Fig.”
Exploratory Factor Analysis With Small Sample Sizes (2009, methodology): “…the correct number of factors in the first series of simulations by using Velicer’s Minimum Average Partial (O’Connor, 2000), a Bayesian Information Criterion (Hansen, Larsen, & Kolenda, 2001), an automatic scree test (Zhu & Ghodsi, 2006), and parallel analysis (O’Connor, 2000; data not shown).”
Fast UniFrac: Facilitating high-throughput phylogenetic analyses of microbial communities including analysis of pyrosequencing and PhyloChip data (2009, background): “…that allow the dimensions that separate particular groups of environments to be readily identified, and (v) scree plots that help researchers more easily discern the number of important dimensions and thus assist in inferring biological significance in complex datasets (Zhu and Ghodsi, 2006).”
Recovering manifold structure in LLM responses through a joint Euclidean mirror (2026, unknown): “…by a low-dimensional manifold it is recommended to look for an elbow in the scree plot of the matrix of empirical pairwise distances ˆ∆ m,n after double centering, plotted in Figure 6, following existing criteria for latent dimensionality selection (see, for example Zhu and Ghodsi, 2006).”
Detecting Perspective Shifts in Multi-agent Systems (2025, unknown): “The number of components for TDKPS was selected automatically via elbow selection (Chung et al., 2019; Zhu & Ghodsi, 2006).”
Asymptotically perfect seeded graph matching without edge correlation (and applications to inference) (2025, unknown): “In practice, we select d using an automatic dimensionality selection method (locating an appropriate elbow of the SCREE plot) based on [63], [64].”
Toward a digital twin of U.S. Congress (2025, unknown): “The histograms are the 1-d FLD projections of the Tweets learned after MDS into d dimensions, where d is data-dependent (34).”
Statistical inference on black-box generative models in the data kernel perspective space (2024, unknown): “We use the profile likelihood of the singular values of D to determine the dimensionality of the DKPS (Zhu and Ghodsi, 2006) and note that this may be larger than two.”
Query-efficient model evaluation using cached responses (2026, unknown): “We fix d = 8 for consistency across experiments, though adaptive methods such as selecting d based on the elbows in the scree plot (Zhu & Ghodsi, 2006) for each task or subtask may yield further improvements.”