Machine learning models are famously vulnerable to adversarial attacks: small ad-hoc perturbations of the data that can catastrophically alter the model predictions.

However,in manyapplications there are several different parameters one might wish to vary: for example, scale and density. In contrast to the one-parameter setting, techniques for applying statistics and machine learning in the setting of multiparameter persistence are not well understood due to the lack of a concise representationoftheresults.

Collaborative inference leverages diverse features provided by different agents (e.g.,sensors)formoreaccurateinference. Acommonsetupiswhereeachagent sends its embedded features instead of the raw data to the Fusion Center (FC) for joint prediction. In this setting, we consider inference phase attacks when asmall fraction of agents is compromised.

This hasbeen shown to be insufficient for generating optimal representation for classification, and to find conditionally invariant representations, usually strong assumptions are needed.

Extensive experiments and analyses have been conducted to characterize and demonstrate the effectiveness of the proposed DivOE.

In all cases, HypGRU achieves the best results when the data are projected to hyperbolic spaces before theyare fed to the network, and all its layers are based on hyperbolic geometry. Results of these networks are obtained using their official code.3,4 We also evaluate a light version of Shift-GCN referred to as Shift-GCN-light, where the numbers of inputand output channels for the input and residual blocks arereduced byafactor of2(thenumber ofinput channels fortheinput block is3). We can also see that whenM = 3, GyroAI-HAUNet outperforms Shift-GCN-light on all the datasets. Overall, whenM = 3, GyroAI-HAUNet is competitive to the best GNN model with far fewer parameters.