A Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials

Abstract

Nanoporous materials (NPMs) selectively adsorb and concentrate gases into their pores, and thus could be used to store, capture, and sense many different gases. Modularly synthesized classes of NPMs, such as covalent organic frameworks (COFs), offer a large number of candidate structures for each adsorption task. A complete NPM-property table, containing measurements of the relevant adsorption properties in the candidate NPMs, would enable the matching of NPMs with adsorption tasks. However, in practice the NPM-property matrix is only partially observed (incomplete); (i) many properties of any given NPM have not been measured and (ii) any given property has not been measured for all NPMs.<br><br>The idea in this work is to leverage the observed (NPM, property) values to impute the missing ones. Similarly, commercial recommendation systems impute missing entries in an incomplete item-customer ratings matrix to recommend items to customers. We demonstrate a COF recommendation system to match COFs with adsorption tasks by training a low rank model of an incomplete COF--adsorption-property matrix. A low rank model, trained on the observed (COF, adsorption property) values, provides (i) predictions of the missing (COF, adsorption property) values and (ii) a "map" of COFs, wherein COFs with similar (dissimilar) adsorption properties congregate (separate). We find the performance of the COF recommendation system varies for different adsorption tasks and diminishes precipitously when the fraction of missing entries exceeds 60%. The concepts in our COF recommendation system can be applied broadly to many different materials and properties. <br>

Design and Implementation of Attention-Based CR System in the Context of Big Data

“…The authors conducted a series of experiments using real-world datasets and the results demonstrated that the proposed method outperforms the baseline method in terms of rating prediction and ranking performance [11]. Sturluson A designed a recommendation system based on covalent organic frameworks (COF), and experimentally demonstrated that this COF recommendation system was able to reasonably rank COF according to most of the adsorption performance metrics, but the predictive effectiveness of the system decreased dramatically when the percentage of missing entries exceeded 60% [12].…”

“…The original input is a high-dimensional sparse one-hot code, which is converted into a low-dimensional dense vector by the embedding layer for network training. The output ( ) 0 a of the embedding layer can be described as in equation (3) [19]. a into the DNN network, the forward propagation process of the DNN network can be described in equation ( 4):…”