All Laplacian score values equal to one

I am working with acoustic emissions signals, and have several signal features stored in each column of a matrix (matrix attached below). I am trying to calculate the Laplacian scores of some features, to help select the most "important" features. I'm using the following code:

[idx score] = fsulaplacian(matrix);   %rows = observations; columns = features

However, the Laplacian score for each feature (i.e. for each column of values) is resulting in a value of one. The typical Laplacian scores should vary from 0 to 1 (typically above 0.7 for these cases), but never seen all of the features being equal to one. Can anyone help?

Thanks in advance!

Christian

Answer 1

[Rohit]

Hi Christian,

I understand that you are using the "fsulaplacian" function in MATLAB to calculate the Laplacian scores of your signal features. However, all the computed scores are consistently equal to one, which is not expected as Laplacian scores typically vary between 0 and 1.

There could be several reasons why you are encountering this issue. Here are a few suggestions to help you troubleshoot:

1. Data normalization: Ensure that you have properly normalized your data before calculating the Laplacian scores. It is important to scale the features so that they have similar ranges, as the Laplacian score calculation can be affected by the magnitude of the features.
2. Data preprocessing: Check if there are any missing values or outliers in your data. These irregularities can affect the Laplacian score calculation. Consider performing data cleaning techniques such as imputation or removing outliers before calculating the scores.
3. Insufficient data variation: If all the features in your matrix have similar distributions or limited variation, it can lead to equal Laplacian scores. Consider exploring your data to understand its distribution and assess if there is enough diversity among the features.
4. Data quality and relevance: Check if the features you are using are truly informative and relevant for your task. It's possible that the Laplacian scores are indicating that the features do not provide much discriminatory power and therefore have similar importance.

You can examine the relationship between your features by utilizing descriptive statistics and visualization techniques. Here are some documentation links you can refer to for more information:

https://in.mathworks.com/help/matlab/descriptive-statistics.html https://in.mathworks.com/help/matlab/ref/matlab.graphics.chart.primitive.histogram.html , https://in.mathworks.com/help/matlab/ref/scatter.html

[Christian Stewart]

Hi Rohit, thanks for your answer - will try out these suggestions and will get back to you. Best, Christian