Full interpretable machine learning in 2D with inline coordinates

Abstract

This paper proposed a new methodology for machine learning in 2-dimensional space (2-D ML) in inline coordinates. It is a full machine learning approach that does not require to deal with n-dimensional data in n-dimensional space. It allows discovering n-D patterns in 2-D space without loss of n-D information using graph representations of n-D data in 2-D. Specifically, it can be done with the inline based coordinates in different modifications, including static and dynamic ones. The classification and regression algorithms based on these inline coordinates were introduced. A successful case study based on a benchmark data demonstrated the feasibility of the approach. This approach helps to consolidate further a whole new area of full 2-D machine learning as a promising ML methodology. It has advantages of abilities to involve actively the end-users into the discovering of models and their justification. Another advantage is providing interpretable ML models.

Visual Knowledge Discovery with Artificial Intelligence: Challenges and Future Directions

“…This creates a new paradigm of visual knowledge discovery in this 2-D/3-D visualization space that we call visual space for short. This paradigm is discussed in Sections 3.4 and 4 below, and in Kovalerchuk and Phan [2021].…”

“…The other process in this area requires overcoming several limitations, such as human understanding of complex multidimensional data and models without "downgrading" it to human perceptual and cognitive limits. Existing methods often lead to the loss of interpretable information, occlusion, clutter and result in quasi-explanations Kovalerchuk et al [2021].…”

“…The user cannot trust them and hardly will have solid, well-grounded further steps of VA such as decision making. Traditional visualization methods, which convert n-D data to 2-D data, are lossy, not preserving all multidimensional information Kovalerchuk [2018], Kovalerchuk et al [2021]. In contrast, the representation of n-D data using General Line Coordinates (GLC) is lossless Kovalerchuk [2018].…”

“…A similar idea is presented in the 2020 review Eisler and Meyer [2020] with visualization for existing ML models. The most recent review from 2021 on visual methods in ML model interpretation Kovalerchuk et al [2021] pointed out that the visualization methods used in DL are very limited. Typically they are heatmaps for activations and bar charts for feature importance.…”

“…Problem 1 Developing new lossless visualizations of multidimensional data. Several types of General Line Coordinates Kovalerchuk [2018] are developed, explored, and applied at different levels (parallel coordinates, radial coordinates, shifted paired coordinates, collocated pairs coordinates, in-line coordinates, GLC-L) Kovalerchuk [2018], Kovalerchuk et al [2021], Kovalerchuk and Phan [2021], Dovhalets et al [2018], , Kovalerchuk and Gharawi [2018], McDonald and Kovalerchuk [2020], Wagle and Kovalerchuk [2020]. Many other GLCs are just defined in Kovalerchuk [2018].…”