Unveil the unseen: Exploit information hidden in noise

Abstract

Noise and uncertainty are usually the enemy of machine learning, noise in training data leads to uncertainty and inaccuracy in the predictions. However, we develop a machine learning architecture that extracts crucial information out of the noise itself to improve the predictions. The phenomenology computes and then utilizes uncertainty in one target variable to predict a second target variable. We apply this formalism to PbZr0.7Sn0.3O3 crystal, using the uncertainty in dielectric constant to extrapolate heat capacity, correctly predicting a phase transition that otherwise cannot be extrapolated. For the second example – single-particle diffraction of droplets – we utilize the particle count together with its uncertainty to extrapolate the ground truth diffraction amplitude, delivering better predictions than when we utilize only the particle count. Our generic formalism enables the exploitation of uncertainty in machine learning, which has a broad range of applications in the physical sciences and beyond.

Predicting the efficiency of luminescent solar concentrators for solar energy harvesting using machine learning

“…Though the models demonstrate to be equally competitive, the Random Forest (RF) regressor (with removed outliers) slightly outperforms the other models. Similar results were reported, using RF to estimate the data uncertainty 32 .…”

Probabilistic selection and design of concrete using machine learning

“…Alternative to the conventional single-layer approach, the flow of information through the two-layer random forest model is demonstrated in Figure 2c. In the two-layer method, we train the first random forest model on the concrete mix proportions (Flow C1) to predict all of the intermediate and output variables such as carbonation coefficient, alongside its uncertainty (Flow C2), following the approach described in Zviazhynski and Conduit (2022). We then train the second layer random forest model, taking the concrete mix proportions (Flow C3) and other recently predicted variables including values for uncertainty (Flow C4) to predict the outputs, for example, strength (Flow C5).…”

“…For example, the use of uncertainty has been extensively demonstrated and experimentally verified for design of materials most likely to fulfill target criteria (Conduit et al, 2017(Conduit et al, , 2018(Conduit et al, , 2019. Furthermore, values of uncertainty itself can be useful for predicting the quantity of interest (Goujon, 2009;Zerva et al, 2017;Zhang, 2020;Zviazhynski and Conduit, 2022). In concrete, the appearance of randomly distributed aggregates of different shapes and sizes may be considered similar to white noise, leading to variability and uncertainty in properties such as carbonation depth and compressive strength.…”