Table 1 This table provides an overview of the key elements extracted from the reviewed research papers. The columns include the reference for each paper, the algorithm employed, the techniques utilized, the main findings, identified gaps, and the data used in the study. This summary aids in understanding the scope, methodologies, and outcomes of the existing research while highlighting areas for future investigation.

1

Wadstromer et al.²⁷

Stacked AE

Natural Scenes

Mean Squared Error

SAE returns small Mean Squared Error

Not studied in terms of classification performance, Execution time not computed, Overlooked class imbalance problem

2

Stephen et al.²⁸

Variational AE

Astronomical, Sloan Digital Sky Survey (SDSS)

Mean Squared Error

Small latent dimension retains enough information

Lacks classification performance analysis, execution time measurement, and consideration of class imbalance.

3

Pandey et al.²⁹

Attention based 1D convolutional AE

Land Cover, Indian Pines

Mean Squared Error, Accuracy

Slightly better than Naïve CNN and RNN based AE

Execution time not computed, Overlooked class imbalance problem

4

Pandey et al.³⁰

Residual Blocks in AE

Indian Pines, Salinas

Mean Squared Error, Accuracy

Good classification Accuracy

Execution time not computed, Overlooked class imbalance problem

5

Pandey et al.³¹

Bidirectional RNN to avoid vanishing gradient in long sequence

Indian Pines, Salinas

Mean Squared Error, Accuracy

Good classification Accuracy

Execution time not computed, Overlooked class imbalance problem

6

Pandey et al.³²

Self-Updating Loops, Feedback loops from deeper layers

Indian Pines

Mean Squared Error, Accuracy

Good classification Accuracy

Execution Time Not Computed, Overlooked class imbalance problem

7

Haut et al.³³

Extreme Learning Machines Based Reduction

Land Cover, Indian Pines, Pavia University, Pavia Center, Houston.

Time, Mean Square Error

Proposed method take more time than PCA but returns better accuracy.

Not studied in terms of classification performance, Overlooked class imbalance problem

9

Fejjari et al.³⁶

Comparison of different Feature extraction techniques

Indian Pines, Salinas

OA, AA, Kappa, Computing time

Linear model are efficient but Non-Linear models return better classification accuracy

Overlooked class imbalance problem

10

Swain et al.³⁷

Linear Models for dimensionality reduction and a hybrid CNN for classification

Indian Pines, Pavia, Salinas

OA, AA, Kappa

Linear Models combined with hybrid CNN returns better classification accuracy

Execution time not computed, Overlooked class imbalance problem

11

Bai et al.⁴³

Two stage SAE, First SAE reduces spectral dimensionality and the second one learns the spectral-spatial features from reduced dimensionality.

Indian Pines, KSC, Salinas

OA

Better classification Accuracy with small number of training samples

Execution time not computed,

Overlooked class imbalance problem