Aardvark Weather: end-to-end data-driven weather forecasting

Machine learning is revolutionising medium-range weather prediction. However it has only been applied to specific and individual components of the weather prediction pipeline. Consequently these data-driven approaches are unable to be deployed without input from conventional operational numerical weather prediction (NWP) systems, which is computationally costly and does not support end-to-end optimisation. In this work, we take a radically different approach and replace the entire NWP pipeline with a machine learning model. We present Aardvark Weather, the first end-to-end data-driven forecasting system which takes raw observations as input and provides both global and local forecasts. These global forecasts are produced for 24 variables at multiple pressure levels at one-degree spatial resolution and 24 hour temporal resolution, and are skillful with respect to hourly climatology at five to seven day lead times. Local forecasts are produced for temperature, mean sea level pressure, and wind speed at a geographically diverse set of weather stations, and are skillful with respect to an IFS-HRES interpolation baseline at multiple lead-times. Aardvark, by virtue of its simplicity and scalability, opens the door to a new paradigm for performing accurate and efficient data-driven medium-range weather forecasting.

Autoregressive Conditional Neural Processes

Conditional neural processes (CNPs; Garnelo et al., 2018a) are attractive meta-learning models which produce well-calibrated predictions and are trainable via a simple maximum likelihood procedure. Although CNPs have many advantages, they are unable to model dependencies in their predictions. Various works propose solutions to this, but these come at the cost of either requiring approximate inference or being limited to Gaussian predictions. In this work, we instead propose to change how CNPs are deployed at test time, without any modifications to the model or training procedure. Instead of making predictions independently for every target point, we autoregressively define a joint predictive distribution using the chain rule of probability, taking inspiration from the neural autoregressive density estimator (NADE) literature. We show that this simple procedure allows factorised Gaussian CNPs to model highly dependent, non-Gaussian predictive distributions. Perhaps surprisingly, in an extensive range of tasks with synthetic and real data, we show that CNPs in autoregressive (AR) mode not only significantly outperform non-AR CNPs, but are also competitive with more sophisticated models that are significantly more computationally expensive and challenging to train. This performance is remarkable given that AR CNPs are not trained to model joint dependencies. Our work provides an example of how ideas from neural distribution estimation can benefit neural processes, and motivates research into the AR deployment of other neural process models.

Active Learning with Convolutional Gaussian Neural Processes for Environmental Sensor Placement

Deploying environmental measurement stations can be a costly and time-consuming procedure, especially in remote regions that are difficult to access, such as Antarctica. Therefore, it is crucial that sensors are placed as efficiently as possible, maximising the informativeness of their measurements. This can be tackled by fitting a probabilistic model to existing data and identifying placements that would maximally reduce the model's uncertainty. The models most widely used for this purpose are Gaussian processes (GPs). However, designing a GP covariance which captures the complex behaviour of non-stationary spatiotemporal data is a difficult task. Further, the computational cost of GPs makes them challenging to scale to large environmental datasets. In this work, we explore using a convolutional Gaussian neural process (ConvGNP) to address these issues. A ConvGNP is a meta-learning model that uses neural networks to parameterise a GP predictive. Our model is data-driven, flexible, efficient, and permits multiple input predictors of gridded or scattered modalities. Using simulated surface air temperature fields over Antarctica as ground truth, we show that a ConvGNP significantly outperforms a non-stationary GP baseline in terms of predictive performance. We then use the ConvGNP in an Antarctic sensor placement toy experiment, yielding promising results.

Sparse Gaussian Process Hyperparameters: Optimize or Integrate?

The kernel function and its hyperparameters are the central model selection choice in a Gaussian proces (Rasmussen and Williams, 2006). Typically, the hyperparameters of the kernel are chosen by maximising the marginal likelihood, an approach known as Type-II maximum likelihood (ML-II). However, ML-II does not account for hyperparameter uncertainty, and it is well-known that this can lead to severely biased estimates and an underestimation of predictive uncertainty. While there are several works which employ a fully Bayesian characterisation of GPs, relatively few propose such approaches for the sparse GPs paradigm. In this work we propose an algorithm for sparse Gaussian process regression which leverages MCMC to sample from the hyperparameter posterior within the variational inducing point framework of Titsias (2009). This work is closely related to Hensman et al. (2015b) but side-steps the need to sample the inducing points, thereby significantly improving sampling efficiency in the Gaussian likelihood case. We compare this scheme against natural baselines in literature along with stochastic variational GPs (SVGPs) along with an extensive computational analysis.

A Note on the Chernoff Bound for Random Variables in the Unit Interval

The Chernoff bound is a well-known tool for obtaining a high probability bound on the expectation of a Bernoulli random variable in terms of its sample average. This bound is commonly used in statistical learning theory to upper bound the generalisation risk of a hypothesis in terms of its empirical risk on held-out data, for the case of a binary-valued loss function. However, the extension of this bound to the case of random variables taking values in the unit interval is less well known in the community. In this note we provide a proof of this extension for convenience and future reference.

Practical Conditional Neural Processes Via Tractable Dependent Predictions

Conditional Neural Processes (CNPs; Garnelo et al., 2018a) are meta-learning models which leverage the flexibility of deep learning to produce well-calibrated predictions and naturally handle off-the-grid and missing data. CNPs scale to large datasets and train with ease. Due to these features, CNPs appear well-suited to tasks from environmental sciences or healthcare. Unfortunately, CNPs do not produce correlated predictions, making them fundamentally inappropriate for many estimation and decision making tasks. Predicting heat waves or floods, for example, requires modelling dependencies in temperature or precipitation over time and space. Existing approaches which model output dependencies, such as Neural Processes (NPs; Garnelo et al., 2018b) or the FullConvGNP (Bruinsma et al., 2021), are either complicated to train or prohibitively expensive. What is needed is an approach which provides dependent predictions, but is simple to train and computationally tractable. In this work, we present a new class of Neural Process models that make correlated predictions and support exact maximum likelihood training that is simple and scalable. We extend the proposed models by using invertible output transformations, to capture non-Gaussian output distributions. Our models can be used in downstream estimation tasks which require dependent function samples. By accounting for output dependencies, our models show improved predictive performance on a range of experiments with synthetic and real data.

Modelling Non-Smooth Signals with Complex Spectral Structure

The Gaussian Process Convolution Model (GPCM; Tobar et al., 2015a) is a model for signals with complex spectral structure. A significant limitation of the GPCM is that it assumes a rapidly decaying spectrum: it can only model smooth signals. Moreover, inference in the GPCM currently requires (1) a mean-field assumption, resulting in poorly calibrated uncertainties, and (2) a tedious variational optimisation of large covariance matrices. We redesign the GPCM model to induce a richer distribution over the spectrum with relaxed assumptions about smoothness: the Causal Gaussian Process Convolution Model (CGPCM) introduces a causality assumption into the GPCM, and the Rough Gaussian Process Convolution Model (RGPCM) can be interpreted as a Bayesian nonparametric generalisation of the fractional Ornstein-Uhlenbeck process. We also propose a more effective variational inference scheme, going beyond the mean-field assumption: we design a Gibbs sampler which directly samples from the optimal variational solution, circumventing any variational optimisation entirely. The proposed variations of the GPCM are validated in experiments on synthetic and real-world data, showing promising results.

Wide Mean-Field Bayesian Neural Networks Ignore the Data

Bayesian neural networks (BNNs) combine the expressive power of deep learning with the advantages of Bayesian formalism. In recent years, the analysis of wide, deep BNNs has provided theoretical insight into their priors and posteriors. However, we have no analogous insight into their posteriors under approximate inference. In this work, we show that mean-field variational inference entirely fails to model the data when the network width is large and the activation function is odd. Specifically, for fully-connected BNNs with odd activation functions and a homoscedastic Gaussian likelihood, we show that the optimal mean-field variational posterior predictive (i.e., function space) distribution converges to the prior predictive distribution as the width tends to infinity. We generalize aspects of this result to other likelihoods. Our theoretical results are suggestive of underfitting behavior previously observered in BNNs. While our convergence bounds are non-asymptotic and constants in our analysis can be computed, they are currently too loose to be applicable in standard training regimes. Finally, we show that the optimal approximate posterior need not tend to the prior if the activation function is not odd, showing that our statements cannot be generalized arbitrarily.

How Tight Can PAC-Bayes be in the Small Data Regime?

In this paper, we investigate the question: Given a small number of datapoints, for example N = 30, how tight can PAC-Bayes and test set bounds be made? For such small datasets, test set bounds adversely affect generalisation performance by discarding data. In this setting, PAC-Bayes bounds are especially attractive, due to their ability to use all the data to simultaneously learn a posterior and bound its generalisation risk. We focus on the case of i.i.d. data with a bounded loss and consider the generic PAC-Bayes theorem of Germain et al. (2009) and Begin et al. (2016). While their theorem is known to recover many existing PAC-Bayes bounds, it is unclear what the tightest bound derivable from their framework is. Surprisingly, we show that for a fixed learning algorithm and dataset, the tightest bound of this form coincides with the tightest bound of the more restrictive family of bounds considered in Catoni (2007). In contrast, in the more natural case of distributions over datasets, we give examples (both analytic and numerical) showing that the family of bounds in Catoni (2007) can be suboptimal. Within the proof framework of Germain et al. (2009) and Begin et al. (2016), we establish a lower bound on the best bound achievable in expectation, which recovers the Chernoff test set bound in the case when the posterior is equal to the prior. Finally, to illustrate how tight these bounds can potentially be, we study a synthetic one-dimensional classification task in which it is feasible to meta-learn both the prior and the form of the bound to obtain the tightest PAC-Bayes and test set bounds possible. We find that in this simple, controlled scenario, PAC-Bayes bounds are surprisingly competitive with comparable, commonly used Chernoff test set bounds. However, the sharpest test set bounds still lead to better guarantees on the generalisation error than the PAC-Bayes bounds we consider.