Generalized Unsupervised Manifold Alignment

Authors: Zhen Cui, Hong Chang, Shiguang Shan, Xilin Chen

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on dataset matching and real-world applications demonstrate the effectiveness and the practicability of our manifold alignment method. To validate the effectiveness of the proposed manifold alignment method, we first conduct two manifold alignment experiments on dataset matching, including the alignment of face image sets across different appearance variations and structure matching of protein sequences. Further applications are also performed on video face recognition and visual domain adaptation to demonstrate the practicability of the proposed method.
Researcher Affiliation Academia Key Lab of Intelligent Information Processing of Chinese Academy of Sciences (CAS), Institute of Computing Technology, CAS, Beijing, China. School of Computer Science and Technology, Huaqiao University, Xiamen, China.
Pseudocode Yes Algorithm 1 Manifold alignment
Open Source Code No The paper does not provide concrete access to source code for the methodology described, nor does it explicitly state that the code is publicly available.
Open Datasets Yes Here we use Multi-PIE database [13]. Here we use the recent published You Tube faces dataset [32]. We use four public datasets, Amazon, Webcam, and DSLR collected in [24], and Caltech-256 [12].
Dataset Splits No The paper does not explicitly provide distinct validation dataset splits for hyperparameter tuning or model selection.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions various algorithms and methods (e.g., PCA, CCA, NN classifier, SURF) but does not provide specific version numbers for any software libraries or dependencies used.
Experiment Setup Yes The main parameters of our method are the balance parameters γf, γp, which are simply set to 1. In the geometry preserving term, we set the nearest neighbor number K = 5 and the heat kernel parameter to 1. The embedding dimension d is set to the minimal rank of two sets minus 1.