Search for: All records

This paper presents an active distillation method for a local institution (e.g., hospital) to find the best queries within its given budget to distill an on-server black-box model’s predictive knowledge into a local surrogate with transparent parameterization. This allows local institutions to understand better the predictive reasoning of the black-box model in its own local context or to further customize the distilled knowledge with its private dataset that cannot be centralized and fed into the server model. The proposed method thus addresses several challenges of deploying machine learning (ML) in many industrial settings (e.g., healthcare analytics) with strong proprietary constraints. These include: (1) the opaqueness of the server model’s architecture which prevents local users from understanding its predictive reasoning in their local data contexts; (2) the increasing cost and risk of uploading local data on the cloud for analysis; and (3) the need to customize the server model with private onsite data. We evaluated the proposed method on both benchmark and real-world healthcare data where significant improvements over existing local distillation methods were observed. A theoretical analysis of the proposed method is also presented. 

Efforts to reduce the carbon footprint associated with cement and concrete production have resulted in a number of promising lower‐emission alternatives. Still, research has emphasized a small subset of potentially useful precursor materials. With the goal of expanding the precursor pool, this work presents results of parallel literature mining and rate modeling activities. As a result of literature mining, materials with appropriate SiO₂, Al₂O₃, and CaO concentrations were assembled into a comprehensive, representative ternary diagram. 23 000+ materials were extracted from 7000 journal articles, and 7500 materials from 6000 articles with 80 ≤ SiO₂ + Al₂O₃ + CaO ≤105 wt% automatically classified. Both supervised and semi‐supervised models were used for dissolution rate prediction of glassy materials with all models pulling from a single data set (n = 802 reported dissolution rates from 105 different glasses). Supervised modeling utilized linear and decision tree regressions to determine features most predictive of dissolution rate, resulting in log‐linear relationships between rate and pH, inverse temperature (1/K), and non‐bridging oxygen per tetrahedron (NBO/T). Semi‐supervised modeling was observed to be more robust to broader feature inclusion, providing similar predictive ability with a relatively larger set of descriptive features. Most importantly, results indicated that models trained on data from disparate scientific communities were adequately predictive (RMSE ≈ 1), particularly under pH ≥7 conditions relevant to the cement and alkali activation communities.