Deep Hurdle Networks for Zero-Inflated Multi-Target Regression: Application to Multiple Species Abundance Estimation

Kong, Shufeng; Bai, Junwen; Lee, Jae Hee; Chen, Di; Allyn, Andrew; Stuart, Michelle; Pinsky, Malin L.; Mills, Katherine E.; Gomes, Carla P.

doi:10.48550/arxiv.2010.16040

Cited by 4 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Specifically, Mat2Spec's feature encoder was inspired by the GATGNN, 9 CGCNN, 26 and MEGNet 12 models in which GNNs are used to encode the crystal structures. The encoding of the labels and features onto a latent Gaussian mixture space to exploit underlying correlations was inspired by the DMVP, 43,44 DHN, 45 and H-CLMP 7 models in which the latent Gaussian spaces learn multiple properties' correlations. Integrating correlation learning with neural networks was initially motivated by computational sustainability applications.…”

Section: Discussionmentioning

confidence: 99%

Density of States Prediction for Materials Discovery via Contrastive Learning from Probabilistic Embeddings

Kong,

Ricci,

Guevarra

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Machine learning for materials discovery has largely focused on predicting an individual scalar rather than multiple related properties, where spectral properties are an important example. Fundamental spectral properties include the phonon density of states (phDOS) and the electronic density of states (eDOS), which individually or collectively are the origins of a breadth of materials observables and functions. Building upon the success of graph attention networks for encoding crystalline materials, we introduce a probabilistic embedding generator specifically tailored to the prediction of spectral properties. Coupled with supervised contrastive learning, our materials-to-spectrum (Mat2Spec) model outperforms state-of-the-art methods for predicting ab initio phDOS and eDOS for crystalline materials. We demonstrate Mat2Spec's ability to identify eDOS gaps below the Fermi energy, validating predictions with ab initio calculations and thereby discovering candidate thermoelectrics and transparent conductors. Mat2Spec is an exemplar framework for predicting spectral properties of materials via strategically incorporated machine learning techniques.

show abstract

Section: Discussionmentioning

confidence: 99%

Density of States Prediction for Materials Discovery via Contrastive Learning from Probabilistic Embeddings

Kong,

Ricci,

Guevarra

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The integration of AI techniques in the present work build upon the foundation of our prior work developing general multi-label classification and and multitarget regression approaches that were initially motivated by ecology applications. These prior works demonstrated the utility of multivariate Gaussian used for pairwise correlation learning, 35 model alignment with a VAE during training, 36,38 and high-order correlation learning via an attention graph neural network. 37 Our careful crafting of model architecture is particularly motivated by the need to make predictions in new composition spaces.…”

Section: Discussionmentioning

confidence: 99%

“…We tackle this challenge with correlation learning, which has been demonstrated to enhance multi-label classification and multi-target regression. [35][36][37][38] Since the multiple properties being predicted may not be explicitly correlated, we developed a framework to learn correlations in latent embeddings of the multiple properties.…”

Section: Introductionmentioning

confidence: 99%

Materials Representation and Transfer Learning for Multi-Property Prediction

Kong¹,

Guevarra²,

Gomes³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

The adoption of machine learning in materials science has rapidly transformed materials property prediction. Hurdles limiting full capitalization of recent advancements in machine learning include the limited development of methods to learn the underlying interactions of multiple elements, as well as the relationships among multiple properties, to facilitate property prediction in new composition spaces. To address these issues, we introduce the Hierarchical Correlation Learning for Multi-property Prediction (H-CLMP) framework that seamlessly integrates (i) prediction using only a material's composition, (ii) learning and exploitation of correlations among target properties in multi-target regression, and (iii) leveraging training data from tangential domains via generative transfer learning. The model is demonstrated for prediction of spectral optical absorption of complex metal oxides spanning 69 3-cation metal oxide composition spaces. H-CLMP accurately predicts non-linear composition-property relationships in composition spaces for which no training data is available, which broadens the purview of machine learning to the discovery of materials with exceptional properties. This achievement results from the principled integration of latent embedding learning, property correlation learning, generative transfer learning, and attention models. The best performance is obtained using H-CLMP with Transfer learning (H-CLMP(T)) wherein a generative adversarial network is trained on computational density of states data and deployed in the target domain to augment prediction of optical absorption from composition. H-CLMP(T) aggregates multiple knowledge sources with a framework that is well-suited for multi-target regression across the physical sciences. SIGNIFICANCE STATEMENTMachine learning algorithms have largely been developed to automate tasks that are routine for humans, which stands in contrast to scientific discovery where one seeks to predict something that has never been observed. For example, in materials science the chemistry of new combinations of 3 elements can yield properties beyond those observed by combining any 2 of the elements. We present a model based on hierarchical correlation learning that makes some such discoveries by learning chemical interactions from other domains via transfer learning. The work aims to emulate a scientist's aggregation of multiple knowledge sources to make informed predictions in new spaces.

show abstract

“…Two-stage models or class weighting approaches can be used to address this in the single-species context but are not feasible in multi-species contexts due to correlations in abundance (especially zeros) across species. We converted abundances to factors, which reduces the impact of zeros, but multivariate hurdle approaches may work better (Kong et al 2020). Additionally, the structure of errors in abundance space for prioritization suggests that an additional model could be coupled to the prediction model to better approximate the D function (i.e.…”

Section: Conceptual Considerationsmentioning

confidence: 99%

“…However, these methods require longer time series than typically available (Chang et al 2017) as do other machine learning methods (Baranwal et al 2021;Clark et al 2021;Kong et al 2020;Rammer & Seidl 2019), e.g. >37,000 observations for (Civantos-Gómez et al 2021).…”

Section: Introductionmentioning

confidence: 99%

Predicting and prioritizing community assembly: learning outcomes via experiments

Blonder

Godoy

2022

Preprint

View full text Add to dashboard Cite

Predicting species coexistence can be difficult often because underlying assembly processes are unknown and data are limited. However, accurate predictions are needed for design and forecasting problems in biodiversity conservation, climate change, invasion ecology, restoration ecology, and synthetic ecology. Here we describe an approach (Learning Outcomes Via Experiments; LOVE) where a limited set of experiments are conducted and multiple community outcomes measured (richness, composition, and abundance), from which a model is trained to predict outcomes for arbitrary experiments. Across seven taxonomically diverse datasets, LOVE predicts test outcomes with low error when trained on ~100 randomly-selected experiments. LOVE can then prioritize experiments for tasks like maximizing outcome richness or total abundance, or minimizing abundances of unwanted species. LOVE complements existing mechanism-first approaches to prediction and shows that rapid screening of communities for desirable properties may become possible.

show abstract

Deep Hurdle Networks for Zero-Inflated Multi-Target Regression: Application to Multiple Species Abundance Estimation

Cited by 4 publications

References 13 publications

Density of States Prediction for Materials Discovery via Contrastive Learning from Probabilistic Embeddings

Density of States Prediction for Materials Discovery via Contrastive Learning from Probabilistic Embeddings

Materials Representation and Transfer Learning for Multi-Property Prediction

Predicting and prioritizing community assembly: learning outcomes via experiments

Contact Info

Product

Resources

About