Mandana Saebi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Mandana Saebi has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
Nitesh Chawla
4 shared publicationsBozhao Nan
3 shared publicationsJessica Wahlers
3 shared publications- Nitesh V. Chawla
University of Notre Dame, Notre Dame, Indiana, USA
3 shared publications Olaf Wiest
3 shared publicationsDavid M. Lodge
Affiliation Cornell Atkins on Center for Sustainability and Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, New York, United States of America
2 shared publications
Latest Publications
- 1 January 2022
- Published by Elsevier in Journal of Environmental Management
- Vol. 301, 113779
Economic and environmental impacts of ballast water management on Small Island Developing States and Least Developed Countries
- 1 January 2022
- Published by Elsevier in Journal of Environmental Management
- Vol. 301, 113779
Economic and environmental impacts of ballast water management on Small Island Developing States and Least Developed Countries
- Open Access
- 27 September 2021
- Published by Cambridge University Press (CUP) in ChemRxiv
On the Use of Real-World Datasets for Reaction Yield Prediction
The lack of publicly available, large, and unbiased datasets is a key bottleneck for the application of machine learning (ML) methods in synthetic chemistry. Data from electronic laboratory notebooks (ELNs) could provide less biased, large datasets, but no such datasets have been made publicly available. The first real-world dataset from the ELNs of a large pharmaceutical company is disclosed and its relationship to high-throughput experimentation (HTE) datasets is described. For chemical yield predictions, a key task in chemical synthesis, an attributed graph neural network (AGNN) performs as good or better than the best previous models on two HTE datasets for the Suzuki and Buchwald-Hartwig reactions. However, training of the AGNN on the ELN dataset does not lead to a predictive model. The implications of using ELN data for training ML-based models are discussed in the context of yield predictions.
- 27 September 2021
- Published by Cambridge University Press (CUP) in ChemRxiv
Open AccessOn the Use of Real-World Datasets for Reaction Yield Prediction
The lack of publicly available, large, and unbiased datasets is a key bottleneck for the application of machine learning (ML) methods in synthetic chemistry. Data from electronic laboratory notebooks (ELNs) could provide less biased, large datasets, but no such datasets have been made publicly available. The first real-world dataset from the ELNs of a large pharmaceutical company is disclosed and its relationship to high-throughput experimentation (HTE) datasets is described. For chemical yield predictions, a key task in chemical synthesis, an attributed graph neural network (AGNN) performs as good or better than the best previous models on two HTE datasets for the Suzuki and Buchwald-Hartwig reactions. However, training of the AGNN on the ELN dataset does not lead to a predictive model. The implications of using ELN data for training ML-based models are discussed in the context of yield predictions.
- Open Access
- 15 June 2021
- Published by American Chemical Society (ACS) in ChemRxiv
Graph Neural Networks for Predicting Chemical Reaction Performance
Chemical reactions are a complex process, as they involve interaction between several molecular compounds. As a result, predicting the success of a reaction is a non-trivial task, which often requires running several experiments in the lab. This process is is expensive, time consuming, and inefficient. As a result, in recent years, researchers have explored the use of machine learning algorithms to predict reaction success. These methods mainly rely on chemical properties of the molecules involved in the reactions. Despite their promising success, none of existing methods explored the use of structural properties of molecules in predicting reaction success. In this work, we develop an Attributed Graph Neural Network model that integrates both structural properties as well as chemical properties of molecules for predicting reaction success. Our model shows remarkable performance on two hand-crafted datasets obtained from high-throughput experiments, as well as one real-world dataset.
- 15 June 2021
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessGraph Neural Networks for Predicting Chemical Reaction Performance
Chemical reactions are a complex process, as they involve interaction between several molecular compounds. As a result, predicting the success of a reaction is a non-trivial task, which often requires running several experiments in the lab. This process is is expensive, time consuming, and inefficient. As a result, in recent years, researchers have explored the use of machine learning algorithms to predict reaction success. These methods mainly rely on chemical properties of the molecules involved in the reactions. Despite their promising success, none of existing methods explored the use of structural properties of molecules in predicting reaction success. In this work, we develop an Attributed Graph Neural Network model that integrates both structural properties as well as chemical properties of molecules for predicting reaction success. Our model shows remarkable performance on two hand-crafted datasets obtained from high-throughput experiments, as well as one real-world dataset.
- Open Access
- 17 May 2021
- Published by American Chemical Society (ACS) in ChemRxiv
Graph Neural Networks for Predicting Chemical Reaction Performance
Chemical reactions are a complex process, as they involve interaction between several molecular compounds. As a result, predicting the success of a reaction is a non-trivial task, which often requires running several experiments in the lab. This process is is expensive, time consuming, and inefficient. As a result, in recent years, researchers have explored the use of machine learning algorithms to predict reaction success. These methods mainly rely on chemical properties of the molecules involved in the reactions. Despite their promising success, none of existing methods explored the use of structural properties of molecules in predicting reaction success. In this work, we develop an Attributed Graph Neural Network model that integrates both structural properties as well as chemical properties of molecules for predicting reaction success. Our model shows remarkable performance on two hand-crafted datasets obtained from high-throughput experiments, as well as one real-world dataset.
- 17 May 2021
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessGraph Neural Networks for Predicting Chemical Reaction Performance
Chemical reactions are a complex process, as they involve interaction between several molecular compounds. As a result, predicting the success of a reaction is a non-trivial task, which often requires running several experiments in the lab. This process is is expensive, time consuming, and inefficient. As a result, in recent years, researchers have explored the use of machine learning algorithms to predict reaction success. These methods mainly rely on chemical properties of the molecules involved in the reactions. Despite their promising success, none of existing methods explored the use of structural properties of molecules in predicting reaction success. In this work, we develop an Attributed Graph Neural Network model that integrates both structural properties as well as chemical properties of molecules for predicting reaction success. Our model shows remarkable performance on two hand-crafted datasets obtained from high-throughput experiments, as well as one real-world dataset.
- Open Access
HONEM: Learning Embedding for Higher Order Networks
Representation learning on networks offers a powerful alternative to the oft painstaking process of manual feature engineering, and as a result, has enjoyed considerable success in recent years. However, all the existing representation learning methods are based on the first-order network (FON), that is, the network that only captures the pairwise interactions between the nodes. As a result, these methods may fail to incorporate non-Markovian higher-order dependencies in the network. Thus, the embeddings that are generated may not accurately represent of the underlying phenomena in a network, resulting in inferior performance in different inductive or transductive learning tasks. To address this challenge, this paper presents HONEM, a higher-order network embedding method that captures the non-Markovian higher-order dependencies in a network. HONEM is specifically designed for the higher-order network structure (HON) and outperforms other state-of-the-art methods in node classification, network re-construction, link prediction, and visualization for networks that contain non-Markovian higher-order dependencies.
Open AccessHONEM: Learning Embedding for Higher Order Networks
Representation learning on networks offers a powerful alternative to the oft painstaking process of manual feature engineering, and as a result, has enjoyed considerable success in recent years. However, all the existing representation learning methods are based on the first-order network (FON), that is, the network that only captures the pairwise interactions between the nodes. As a result, these methods may fail to incorporate non-Markovian higher-order dependencies in the network. Thus, the embeddings that are generated may not accurately represent of the underlying phenomena in a network, resulting in inferior performance in different inductive or transductive learning tasks. To address this challenge, this paper presents HONEM, a higher-order network embedding method that captures the non-Markovian higher-order dependencies in a network. HONEM is specifically designed for the higher-order network structure (HON) and outperforms other state-of-the-art methods in node classification, network re-construction, link prediction, and visualization for networks that contain non-Markovian higher-order dependencies.
- Open Access
- 31 July 2020
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 15 (7) , e0220353
Higher-order patterns of aquatic species spread through the global shipping network
The introduction and establishment of nonindigenous species (NIS) through global ship movements poses a significant threat to marine ecosystems and economies. While ballast-vectored invasions have been partly addressed by some national policies and an international agreement regulating the concentrations of organisms in ballast water, biofouling-vectored invasions remain largely unaddressed. Development of additional efficient and cost-effective ship-borne NIS policies requires an accurate estimation of NIS spread risk from both ballast water and biofouling. We demonstrate that the first-order Markovian assumption limits accurate modeling of NIS spread risks through the global shipping network. In contrast, we show that higher-order patterns provide more accurate NIS spread risk estimates by revealing indirect pathways of NIS transfer using Species Flow Higher-Order Networks (SF-HON). Using the largest available datasets of non-indigenous species for Europe and the United States, we then compare SF-HON model predictions against those from networks that consider only first-order connections and those that consider all possible indirect connections without consideration of their significance. We show that not only SF-HONs yield more accurate NIS spread risk predictions, but there are important differences in NIS spread via the ballast and biofouling vectors. Our work provides information that policymakers can use to develop more efficient and targeted prevention strategies for ship-borne NIS spread management, especially as management of biofouling is of increasing concern.
- 31 July 2020
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 15 (7) , e0220353
Open AccessHigher-order patterns of aquatic species spread through the global shipping network
The introduction and establishment of nonindigenous species (NIS) through global ship movements poses a significant threat to marine ecosystems and economies. While ballast-vectored invasions have been partly addressed by some national policies and an international agreement regulating the concentrations of organisms in ballast water, biofouling-vectored invasions remain largely unaddressed. Development of additional efficient and cost-effective ship-borne NIS policies requires an accurate estimation of NIS spread risk from both ballast water and biofouling. We demonstrate that the first-order Markovian assumption limits accurate modeling of NIS spread risks through the global shipping network. In contrast, we show that higher-order patterns provide more accurate NIS spread risk estimates by revealing indirect pathways of NIS transfer using Species Flow Higher-Order Networks (SF-HON). Using the largest available datasets of non-indigenous species for Europe and the United States, we then compare SF-HON model predictions against those from networks that consider only first-order connections and those that consider all possible indirect connections without consideration of their significance. We show that not only SF-HONs yield more accurate NIS spread risk predictions, but there are important differences in NIS spread via the ballast and biofouling vectors. Our work provides information that policymakers can use to develop more efficient and targeted prevention strategies for ship-borne NIS spread management, especially as management of biofouling is of increasing concern.
- Open Access
Heterogeneous Relational Reasoning in Knowledge Graphs with Reinforcement Learning
Path-based relational reasoning over knowledge graphs has become increasingly popular due to a variety of downstream applications such as question answering in dialogue systems, fact prediction, and recommender systems. In recent years, reinforcement learning (RL) has provided solutions that are more interpretable and explainable than other deep learning models. However, these solutions still face several challenges, including large action space for the RL agent and accurate representation of entity neighborhood structure. We address these problems by introducing a type-enhanced RL agent that uses the local neighborhood information for efficient path-based reasoning over knowledge graphs. Our solution uses graph neural network (GNN) for encoding the neighborhood information and utilizes entity types to prune the action space. Experiments on real-world dataset show that our method outperforms state-of-the-art RL methods and discovers more novel paths during the training procedure.
Open AccessHeterogeneous Relational Reasoning in Knowledge Graphs with Reinforcement Learning
Path-based relational reasoning over knowledge graphs has become increasingly popular due to a variety of downstream applications such as question answering in dialogue systems, fact prediction, and recommender systems. In recent years, reinforcement learning (RL) has provided solutions that are more interpretable and explainable than other deep learning models. However, these solutions still face several challenges, including large action space for the RL agent and accurate representation of entity neighborhood structure. We address these problems by introducing a type-enhanced RL agent that uses the local neighborhood information for efficient path-based reasoning over knowledge graphs. Our solution uses graph neural network (GNN) for encoding the neighborhood information and utilizes entity types to prune the action space. Experiments on real-world dataset show that our method outperforms state-of-the-art RL methods and discovers more novel paths during the training procedure.
- Open Access
Higher-order patterns of aquatic species spread through the global shipping network
The introduction and establishment of non-indigenous species (NIS) through global ship movements is a significant threat to marine ecosystems and economies. While ballast-vectored invasions have been partly addressed by some national policies and an international agreement regulating the concentrations of organisms in ballast water, biofouling-vectored invasions remain a large risk. Development of additional realistic and cost-effective ship-borne NIS policies requires an accurate estimation of NIS spread risk from both ballast water and biofouling. In this paper, we demonstrate that first-order Markov assumptions limit accurate modeling of NIS spread risks through the global shipping network. In contrast, we show that higher-order patterns overcome this limitation by revealing indirect pathways of NIS transfer. We accomplish this by developing Species Flow Higher-Order Networks (SF-HON), which we developed independently for ballast and biofouling, for comparison with first-order Markovian models of ballast and biofouling. We evaluated SF-HON predictions using the largest available datasets of invasive species for Europe and the United States. We show that not only does SF-HON yield more accurate NIS spread risk predictions than first-order models and existing higher-order models, but also that there are important differences in NIS spread via the ballast and biofouling vectors. Our work provides information that policymakers can use to develop more efficient and targeted prevention strategies for ship-borne NIS spread management, especially as management of biofouling is of increasing concern.
Open AccessHigher-order patterns of aquatic species spread through the global shipping network
The introduction and establishment of non-indigenous species (NIS) through global ship movements is a significant threat to marine ecosystems and economies. While ballast-vectored invasions have been partly addressed by some national policies and an international agreement regulating the concentrations of organisms in ballast water, biofouling-vectored invasions remain a large risk. Development of additional realistic and cost-effective ship-borne NIS policies requires an accurate estimation of NIS spread risk from both ballast water and biofouling. In this paper, we demonstrate that first-order Markov assumptions limit accurate modeling of NIS spread risks through the global shipping network. In contrast, we show that higher-order patterns overcome this limitation by revealing indirect pathways of NIS transfer. We accomplish this by developing Species Flow Higher-Order Networks (SF-HON), which we developed independently for ballast and biofouling, for comparison with first-order Markovian models of ballast and biofouling. We evaluated SF-HON predictions using the largest available datasets of invasive species for Europe and the United States. We show that not only does SF-HON yield more accurate NIS spread risk predictions than first-order models and existing higher-order models, but also that there are important differences in NIS spread via the ballast and biofouling vectors. Our work provides information that policymakers can use to develop more efficient and targeted prevention strategies for ship-borne NIS spread management, especially as management of biofouling is of increasing concern.