Graph condensation has emerged as a promising avenue to enable scalable training of GNNs by compressing the training dataset while preserving essential graph characteristics. Our study uncovers significant shortcomings in current graph condensation techniques. First, the majority of the algorithms paradoxically require training on the full dataset to perform condensation. Second, due to their gradient-emulating approach, these methods require fresh condensation for any change in hyperparameters or GNN architecture, limiting their flexibility and reusability. To address these challenges, we present Bonsai, a novel graph condensation method empowered by the observation that *computation trees* form the fundamental processing units of message-passing GNNs. Bonsai condenses datasets by encoding a careful selection of *exemplar* trees that maximize the representation of all computation trees in the training set. This unique approach imparts Bonsai as the first linear-time, model-agnostic graph condensation algorithm for node classification that outperforms existing baselines across $7$ real-world datasets on accuracy, while being $22$ times faster on average. Bonsai is grounded in rigorous mathematical guarantees on the adopted approximation strategies, making it robust to GNN architectures, datasets, and parameters.

GNNs, like other deep learning models, are data and computation hungry. There is a pressing need to scale training of GNNs on large datasets to enable their usage on low-resource environments. Graph distillation is an effort in that direction with the aim to construct a smaller synthetic training set from the original training data without significantly compromising model performance. While initial efforts are promising, this work is motivated by two key observations: (1) Existing graph distillation algorithms themselves rely on training with the full dataset, which undermines the very premise of graph distillation. (2) The distillation process is specific to the target GNN architecture and hyper-parameters and thus not robust to changes in the modeling pipeline. We circumvent these limitations by designing a distillation algorithm called MIRAGE for graph classification. MIRAGE is built on the insight that a message-passing GNN decomposes the input graph into a multiset of computation trees. Furthermore, the frequency distribution of computation trees is often skewed in nature, enabling us to condense this data into a concise distilled summary. By compressing the computation data itself, as opposed to emulating gradient flows on the original training set—a prevalent approach to date—MIRAGE transforms into an unsupervised and architecture-agnostic distillation algorithm. Extensive benchmarking on real-world datasets underscores MIRAGE’s superiority, showcasing enhanced generalization accuracy, data compression, and distillation efficiency when compared to state-of-the-art baselines.

Modelling spatio-temporal processes on road networks is a task of growing importance. While significant progress has been made on developing spatio-temporal graph neural networks (Gnns), existing works are built upon three assumptions that are not practical on real-world road networks. First, they assume sensing on every node of a road network. In reality, due to budget-constraints or sensor failures, all locations (nodes) may not be equipped with sensors. Second, they assume that sensing history is available at all installed sensors. This is unrealistic as well due to sensor failures, loss of packets during communication, etc. Finally, there is an assumption of static road networks. Connectivity within networks change due to road closures, constructions of new roads, etc. In this work, we develop Frigate to address all these shortcomings. Frigate is powered by a spatio-temporal Gnn that integrates positional, topological, and temporal information into rich inductive node representations. The joint fusion of this diverse information is made feasible through a novel combination of gated Lipschitz embeddings with Lstms. We prove that the proposed Gnn architecture is provably more expressive than message-passing Gnns used in state-of-the-art algorithms. The higher expressivity of Frigate naturally translates to superior empirical performance conducted on real-world network-constrained traffic data. In addition, Frigate is robust to frugal sensor deployment, changes in road network connectivity, and temporal irregularity in sensing.

An accurate prediction of El Niño Southern Oscillation (ENSO) holds the key to produce skillful seasonal weather predictions across the globe. All the statistical and dynamical ENSO models developed in the past four decades face a common problem, spring predictability barrier, which is the sudden drop in the skill of the ENSO prediction when the forecast is initiated before the onset of boreal summer. Recent studies suggest that data-driven machine learning models can overcome the spring predictability barrier. We show that using a convolutional long short-term memory (ConvLSTM) network, the monthly mean Nino3.4 index can be skillfully predicted up to one year ahead. The model was also able to predict strong El Niño cases, such as 1997–1998 and 2015–2016 a year ahead. Our results suggest that the proposed ConvLSTM model has significant skill in multiseasonal weather predictions.

Tropical cyclones (TCs) are the most destructive weather systems that form over the tropical oceans, with about 90 storms forming globally every year. The timely detection and tracking of TCs are important for advanced warning to the affected regions. As these storms form over the open oceans far from the continents, remote sensing plays a crucial role in detecting them. Here we present an automated TC detection from satellite images based on a novel deep learning technique. In this study, we propose a multistaged deep learning framework for the detection of TCs, including, 1) a detector—Mask region-convolutional neural network (R-CNN); 2) a wind speed filter; and 3) a classifier—convolutional neural network (CNN). The hyperparameters of the entire pipeline are optimized to showcase the best performance using Bayesian optimization. Results indicate that the proposed approach yields high precision (97.10%), specificity (97.59%), and accuracy (86.55%) for test images.

The synoptic-scale (3–7 days) variability is a dominant contributor to the Indian summer monsoon (ISM) seasonal precipitation. An accurate prediction of ISM precipitation by dynamical or statistical model remains a challenge. Here, we show that the sea level pressure (SLP) can be used as a proxy to predict the active-break cycle as well as the genesis of low-pressure systems (LPSs), using a deep learning model, namely convolutional long short-term memory (ConvLSTM) networks. The deep learning model is able to reliably predict the daily SLP anomalies over central India and the Bay of Bengal at a lead time of 7 days. As the fluctuations in SLP drive the changes in the strength of the atmospheric circulation, the prediction of SLP anomalies is useful in predicting the intensity of ISM. A comparison of the ConvLSTM predicted SLP with the forecast of a conventional numerical weather prediction model indicates that the deep learning model possesses better skill in capturing the synoptic-scale SLP fluctuations over central India and Bay of Bengal.

This paper proposes an augmented model approach for identification of nonlinear errors-in-variables (EIVs) systems. An EIV model accounts for uncertainties in the observations of both inputs and outputs. As the direct identification of nonlinear functions is difficult, we propose to approximate the nonlinear EIV model using multiple ARX models. To estimate the noise-free input signal, we use a collection of particle filters which run in parallel corresponding to each of the multiple ARX models. The parameters of local models are estimated by applying expectation maximization algorithm, under a maximum likelihood framework, using the input-output data of the nonlinear EIV system. Simulated numerical examples and an experiment study on a multitank system are used to illustrate the efficacy of the proposed approach.

Major impediments in developing models based on the input-output data of an industrial process are the outliers in the output and uncertainties in the inputs. To address this problem, this article proposes a robust identification approach for nonlinear errors-in-variables systems. The t- distribution is employed to model the process data to account for the outliers through the adjustable degrees of freedom. Furthermore, we propose to approximate the nonlinear dynamics of the process using multiple local ARX models and combine them using a softmax function based weighting approach. To deal with parameter uncertainties, the identification problem is casted in the Bayesian framework and posterior distributions of the model parameters are estimated using the variational Bayesian approach, instead of point estimations. A numerical example of continuous fermenter as well as an experiment study on the multitank system is employed to demonstrate potential of the proposed method.