This paper proposes a novel robust Fisher Discriminant Analysis (FDA) for discriminative subspace learning in the presence of outliers. The proposed approach is motivated by the maximum-likelihood perspective on FDA and its connection to Kullback-Leibler (KL) divergence minimization. Within the probabilistic FDA framework, we develop a robust model by adopting the $\alpha $ -divergence as a flexible alternative to the KL divergence. The resulting method induces an adaptive redescending weighting scheme, in which each observation is weighted according to its statistical compatibility with the model, where the robustness level continuously controlled by $\alpha $ . As $\alpha $ decreases from 1, the influence of outliers is progressively suppressed, while classical FDA is recovered at $\alpha = 1$ . Combined with a two-fold iterative optimization procedure, the proposed method mitigates contamination at both the class-modeling stage and the projection-learning stage. We further provide theoretical analysis of the robustness mechanism, convergence, and computational complexity analysis to support the effectiveness and efficiency of the proposed method. Extensive experiments on synthetic data and multiple public image datasets under diverse contamination settings demonstrate that the proposed method consistently outperforms representative robust FDA variants and related approaches.

A long-standing challenge for forecasting from big, spatiotemporal data with nonstationarity is scalability. Here we overcome this by developing a scalable forecasting framework for slope stability and hazard dynamics and what-if-scenario analytics. The framework incorporates recent findings from granular physics and dynamics of precursory failure, and delivers explainable, actionable and timely intelligence on an impending collapse. To demonstrate its performance, we applied the framework to a slope stability radar (SSR) monitoring data from a rock slope in an operational mine. The forecast of surface motion for the entire monitoring domain is used to establish probabilistic outcomes for what-if-scenarios, as well as the likely location, geometry and timing of failure. Next we implement a four-pronged approach that exploits SSR data from a second slope having a different but similarly complex dynamics, in addition to data simulations, to test robustness to false positive and/or negative predictions and reproducibility of forecast performance. In anticipation of synergistic advances in remote sensing and analytics, this effort takes the first steps toward an Augmented Intelligence platform for future-focused slope stability analytics and decision-making in real-time.