Sparse regularization techniques provide novel insights into outcome integration processes

Abstract

By exploiting information that is contained in the spatial arrangement of neural activations, multivariate pattern analysis (MVPA) can detect distributed brain activations which are not accessible by standard univariate analysis. Recent methodological advances in MVPA regularization techniques have made it feasible to produce sparse discriminative whole-brain maps with highly specific patterns. Furthermore, the most recent refinement, the Graph Net, explicitly takes the 3D-structure of fMRI data into account. Here, these advanced classification methods were applied to a large fMRI sample (N=70) in order to gain novel insights into the functional localization of outcome integration processes. While the beneficial effect of differential outcomes is well-studied in trial-and-error learning, outcome integration in the context of instruction-based learning has remained largely unexplored. In order to examine neural processes associated with outcome integration in the context of instruction-based learning, two groups of subjects underwent functional imaging while being presented with either differential or ambiguous outcomes following the execution of varying stimulus-response instructions. While no significant univariate group differences were found in the resulting fMRI dataset, L1-regularized (sparse) classifiers performed significantly above chance and also clearly outperformed the standard L2-regularized (dense) Support Vector Machine on this whole-brain between-subject classification task. Moreover, additional L2-regularization via the Elastic Net and spatial regularization by the Graph Net improved interpretability of discriminative weight maps but were accompanied by reduced classification accuracies. Most importantly, classification based on sparse regularization facilitated the identification of highly specific regions differentially engaged under ambiguous and differential outcome conditions, comprising several prefrontal regions previously associated with probabilistic learning, rule integration and reward processing. Additionally, a detailed post-hoc analysis of these regions revealed that distinct activation dynamics underlay the processing of ambiguous relative to differential outcomes. Together, these results show that L1-regularization can improve classification performance while simultaneously providing highly specific and interpretable discriminative activation patterns.

Multi-Region Neural Representation: A novel model for decoding visual stimuli in human brains

“…As a result, it can mostly generate better performance in comparison with other methods, especially for binary classification problems. Therefore, SVM is generally used in the wide range of studies for creating predictive models [4,3,6,11]. The final goal of this section is employing the L1-regularization SVM [8] method for creating binary classification at the ROIs level, and then combining these classifiers by using the Bagging algorithm [9,10] for generating the MVP final predictive model.…”

“…L1/2 regularized SVM [8,22], the Elastic Net, and the Graph Net [23], for predicting different responses in the human brain. They show that L1-regularization can improve classification performance while simultaneously providing highly specific and interpretable discriminative activation patterns [3]. Osher et al proposed a network (graph) based approach by using anatomical regions of the human brain for representing and classifying the different visual stimuli responses (faces, objects, bodies, scenes) [12].…”

“…As mentioned before, normalizing brain image to the standard space can increase the time and space complexities and decrease the robustness of MVP techniques, especially in voxel-based methods [5]. On the one hand, most of the previous studies [3,4,6,11] preferred to use original data sets instead of the standard version because of the mentioned problem. On the other hand, this mapping can provide a normalized view for combing homogeneous data sets.…”

“…As one of the main techniques in taskbased functional Magnetic Resonance Imaging (fMRI) analysis, Multivariate Pattern (MVP) is a conjunction between neuroscience and computer science, which can extract and decode brain patterns by applying the classification methods [1,2]. Indeed, it can predict patterns of neural activities associated with different cognitive states [3,4] and also can define decision surfaces to distinguish different stimuli for decoding the brain and understanding how it works [5,6]. Analyzing the patterns of visual objects is one of the most interesting topics in MVP classification, which can enable us to understand how brain stores and processes the visual stimuli.…”

“…On the one hand, most of the previous studies manually selected the ROIs. On the other hand, defining wrong ROIs can significantly decrease the performance of MVP methods [3,4]. Another challenge is the cost of brain studies.…”