Too Much Information Is No Information: How Machine Learning and Feature Selection Could Help in Understanding the Motor Control of Pointing

Abstract

The aim of this study was to develop the use of Machine Learning techniques as a means of multivariate analysis in studies of motor control. These studies generate a huge amount of data, the analysis of which continues to be largely univariate. We propose the use of machine learning classification and feature selection as a means of uncovering feature combinations that are altered between conditions. High dimensional electromyograms (EMG) vectors were generated as several arm and trunk muscles were recorded while subjects pointed at various angles above and below the gravity neutral horizontal plane. We used Linear Discriminant Analysis (LDA) to carry out binary classifications between the EMG vectors for pointing at a particular angle, versus pointing at the gravity neutral direction. Classification success provided a composite index of muscular adjustments for various task constraints (in this case, pointing angles). In order to find the combination of features that were significantly altered between task conditions, we conducted a post classification feature selection i.e. investigated which combination of features had allowed for the classification. Feature selection was done by comparing the representations of each category created by LDA for the classification. In other words computing the difference between the representations of each class. We propose that this approach will help with comparing high dimensional EMG patterns in two ways; i) quantifying the effects of the entire pattern rather than using single arbitrarily defined variables and ii) identifying the parts of the patterns that convey the most information regarding the investigated effects.

Deactivation and collective phasic muscular tuning for pointing direction: Insights from machine learning

“…The main machine learning technique used for the study was LDA. This technique was primarily chosen because of its simplicity and ease of understanding the features which are critical to the classification [ 25 ]. As a technique which is not as powerful as methods which were developed later [ 20 , [38] , [39] , [40] , [41] ], there was also a lower risk of classification saturation (see Supplementary Fig.…”

“…Having all the elements needed, can then be determined for any new observation using Equation (1) . Additionally, as we did in Thomas et al [ 25 ], we computed the as a further indicator of data separation. We defined this value as the Euclidean distance between the means of the Gaussian distributions representing the classes…”

“…The former algorithm utilizes information on the mean and variance of each category to build a representation of the group [ [22] , [23] , [24] ]). One of the reasons for choosing LDA is the greater ease for finding the factors which are critical in classification success [ 25 ]. This amounts to finding which segments of EMG activity are most different between the groups being discriminated.…”

“…Algorithms like Support Vector Machines are powerful classifiers, in many instances outperforming the more recent deep neural networks [ [28] , [29] , [30] ]. Nevertheless, they are harder to probe for the combination of features which are the most critical to classification accuracy [ 25 ]. Our previous studies have shown that they can be productively applied to the analysis of EMG data [ 18 , 19 ].…”