Data-driven weather prediction (DDWP) models are increasingly becoming popular for weather forecasting. However, while operational weather forecasts predict a wide variety of weather variables, DDWPs currently forecast a specific set of key prognostic variables. Non-prognostic ("diagnostic") variables are sometimes modeled separately as dependent variables of the prognostic variables (c.f. FourCastNet), or by including the diagnostic variable as a target in the DDWP. However, the cost of training and deploying bespoke models for each diagnostic variable can increase dramatically with more diagnostic variables, and limit the operational use of such models. Likewise, retraining an entire DDWP each time a new diagnostic variable is added is also cost-prohibitive. We present an two-stage approach that allows new diagnostic variables to be added to an end-to-end DDWP model without the expensive retraining. In the first stage, we train an autoencoder that learns to embed prognostic variables into a latent space. In the second stage, the autoencoder is frozen and "downstream" models are trained to predict diagnostic variables using only the latent representations of prognostic variables as input. Our experiments indicate that models trained using the two-stage approach offer accuracy comparable to training bespoke models, while leading to significant reduction in resource utilization during training and inference. This approach allows for new "downstream" models to be developed as needed, without affecting existing models and thus reducing the friction in operationalizing new models.
Data-driven weather prediction models (DDWPs) have made rapid strides in recent years, demonstrating an ability to approximate Numerical Weather Prediction (NWP) models to a high degree of accuracy. The fast, accurate, and low-cost DDWP forecasts make their use in operational forecasting an attractive proposition, however, there remains work to be done in rigorously evaluating DDWPs in a true operational setting. Typically trained and evaluated using ERA5 reanalysis data, DDWPs have been tested only in a simulation, which cannot represent the real world with complete accuracy even if it is of a very high quality. The safe use of DDWPs in operational forecasting requires more thorough "real-world" verification, as well as a careful examination of how DDWPs are currently trained and evaluated. It is worth asking, for instance, how well do the reanalysis datasets, used for training, simulate the real world? With an eye towards climate justice and the uneven availability of weather data: is the simulation equally good for all regions of the world, and would DDWPs exacerbate biases present in the training data? Does a good performance in simulation correspond to good performance in operational settings? In addition to approximating the physics of NWP models, how can ML be uniquely deployed to provide more accurate weather forecasts? As a first step towards answering such questions, we present a robust dataset of in-situ observations derived from the NOAA MADIS program to serve as a benchmark to validate DDWPs in an operational setting. By providing a large corpus of quality-controlled, in-situ observations, this dataset provides a meaningful real-world task that all NWPs and DDWPs can be tested against. We hope that this data can be used not only to rigorously and fairly compare operational weather models but also to spur future research in new directions.
Label propagation is a powerful and flexible semi-supervised learning technique on graphs. Neural networks, on the other hand, have proven track records in many supervised learning tasks. In this work, we propose a training framework with a graph-regularised objective, namely "Neural Graph Machines", that can combine the power of neural networks and label propagation. This work generalises previous literature on graph-augmented training of neural networks, enabling it to be applied to multiple neural architectures (Feed-forward NNs, CNNs and LSTM RNNs) and a wide range of graphs. The new objective allows the neural networks to harness both labeled and unlabeled data by: (a) allowing the network to train using labeled data as in the supervised setting, (b) biasing the network to learn similar hidden representations for neighboring nodes on a graph, in the same vein as label propagation. Such architectures with the proposed objective can be trained efficiently using stochastic gradient descent and scaled to large graphs, with a runtime that is linear in the number of edges. The proposed joint training approach convincingly outperforms many existing methods on a wide range of tasks (multi-label classification on social graphs, news categorization, document classification and semantic intent classification), with multiple forms of graph inputs (including graphs with and without node-level features) and using different types of neural networks.
In this paper we propose and investigate a novel end-to-end method for automatically generating short email responses, called Smart Reply. It generates semantically diverse suggestions that can be used as complete email responses with just one tap on mobile. The system is currently used in Inbox by Gmail and is responsible for assisting with 10% of all mobile responses. It is designed to work at very high throughput and process hundreds of millions of messages daily. The system exploits state-of-the-art, large-scale deep learning. We describe the architecture of the system as well as the challenges that we faced while building it, like response diversity and scalability. We also introduce a new method for semantic clustering of user-generated content that requires only a modest amount of explicitly labeled data.