ICA (Independent Component Analysis) compression

Step 1. Start

Select a compression method: Edit → Compress → ICA compression.

Step 2: ICA Input

The function prompts the user to enter the number of independent components to be extracted (used as a measure of how much to compress the data, the default value is 10 components) and the regularization parameter lambda. The default value is 1. The regularization parameter is used in the ICA algorithm to control the trade-off between data fit and complexity of the model.

After completion, IDCube calculates the compression ratio by comparing the size of the original data and the size of the compressed data. This information, along with the number of independent components and the size of the saved file, is then displayed to the user.

The message box indicates that the new compressed datacube file can be opened in a usual way.

Additional Information:

Compared to PCA, which only looks for uncorrelated factors in the data, ICA goes a step further to find independent factors. While uncorrelated components can be statistically related, independent components are not. PCA tends to find the axes in the data that account for the most variance, but it does not necessarily find the actual source signals.

ICA assumes that the signal sources are statistically independent, while PCA does not require this assumption. Because of this, ICA can often provide a more meaningful decomposition of complex datasets and is often used when we expect some underlying factors are driving the patterns in the data.

The actual compression in this case does not involve any physical compression or data encoding. Rather, the datacube is compressed in the sense that it is represented by a smaller set of independent components (endmembers) and their abundances. This is an example of data reduction rather than data compression in the traditional sense.