Figures & data
Figure 2. 3D to 2D dimensionality reduction process executed by a manifold learning method. Adapted from.[Citation51].
![Figure 2. 3D to 2D dimensionality reduction process executed by a manifold learning method. Adapted from.[Citation51].](/cms/asset/02d64571-9aa1-4cfb-9377-49aa26860df7/ljfp_a_2161571_f0002_oc.jpg)
Figure 6. Voltammograms obtained in measurement with the 3 voltammetric sensors of platinum, gold, and carbon working electrodes.
![Figure 6. Voltammograms obtained in measurement with the 3 voltammetric sensors of platinum, gold, and carbon working electrodes.](/cms/asset/c428efeb-f801-48f0-a3de-3c3915e681a5/ljfp_a_2161571_f0006_oc.jpg)
Figure 9. Classification accuracy results when applying PCA as dimensionality reduction method and after performing the LOOCV cross-validation process with the k-NN algorithm.
![Figure 9. Classification accuracy results when applying PCA as dimensionality reduction method and after performing the LOOCV cross-validation process with the k-NN algorithm.](/cms/asset/deb1922c-54ec-4819-aba4-f2f3ba943950/ljfp_a_2161571_f0009_oc.jpg)
Figure 10. Two-dimensional scatter diagram of the first two principal components after executing the dimensionality reduction with the PCA algorithm.
![Figure 10. Two-dimensional scatter diagram of the first two principal components after executing the dimensionality reduction with the PCA algorithm.](/cms/asset/0cd5aa74-e2f1-422e-86d0-8b6577251edb/ljfp_a_2161571_f0010_oc.jpg)
Figure 11. Classification accuracy behavior when parameter changing in each manifold learning algorithm used to perform the dimensionality reduction process. a) Graph k parameter in Laplacian Eigenmaps, b) Graph k parameter in LLE, c) Graph k parameter in Isomap and d) Perplexity p parameter in t-SNE.
![Figure 11. Classification accuracy behavior when parameter changing in each manifold learning algorithm used to perform the dimensionality reduction process. a) Graph k parameter in Laplacian Eigenmaps, b) Graph k parameter in LLE, c) Graph k parameter in Isomap and d) Perplexity p parameter in t-SNE.](/cms/asset/3c46c2c2-cbef-4877-a2ea-d3d0c0c0c7e6/ljfp_a_2161571_f0011_oc.jpg)
Figure 12. 2D scatter plots of the four nonlinear dimensionality reduction algorithms. a) Laplacian Eigenmaps, b) LLE, c) Isomap and d) t- SNE.
![Figure 12. 2D scatter plots of the four nonlinear dimensionality reduction algorithms. a) Laplacian Eigenmaps, b) LLE, c) Isomap and d) t- SNE.](/cms/asset/12b11f2d-ddea-42ea-b374-b38e3ff67a2d/ljfp_a_2161571_f0012_oc.jpg)
Figure 13. Accuracy behavior with a variation of the number of dimensions at the input of the k -NN classifier algorithm for a) LLE, b) t-SNE, c) Isomap and d) Laplacian Eigenmaps.
![Figure 13. Accuracy behavior with a variation of the number of dimensions at the input of the k -NN classifier algorithm for a) LLE, b) t-SNE, c) Isomap and d) Laplacian Eigenmaps.](/cms/asset/8145a909-295b-4fd6-b34d-0f889c4bcb10/ljfp_a_2161571_f0013_oc.jpg)
Figure 14. Confusion matrices and accuracy result of classifier k -NN evaluated with LOOCV for a) Laplacian Eigenmaps, b) LLE, c)Isomap ans d) t-SNE.
![Figure 14. Confusion matrices and accuracy result of classifier k -NN evaluated with LOOCV for a) Laplacian Eigenmaps, b) LLE, c)Isomap ans d) t-SNE.](/cms/asset/44c29eaf-451a-40df-83d4-7261c7885fd9/ljfp_a_2161571_f0014_oc.jpg)