Active research in time series classification and forecasting has led to the development of a wide range of machine learning models. For practitioners, the selection of a suitable model among these, along with their hyperparameters, remains a challenging task. While automated machine learning offers approaches for automatic selection of models for a given task, the practical efficacy of these methods is often limited, due to the computational complexity of searching over a large design space and the high dimensionality of time series datasets that poses additional challenges on generalisation quality. To fill this gap, we propose a meta-learning framework that transfers past knowledge from previous searches to recommend an architecture and its hyperparameters; specifically, this framework utilises a joint representation of deep neural architectures and time series datasets, and predicts the performance of neural architectures along with their hyperparameters on time series datasets. Our computational experiments reveal that the configurations proposed by our meta-learned surrogate achieve a performance gain of up to 34% on 4 out of the 8 forecasting datasets we considered and up to 60% on 36 out of 73 of our classification datasets, whilst reducing the computational cost to 10% of that required by the hyperparameter optimisation method HEBO to tune the architectures, showcasing the effectiveness of meta-learning in the time series domain.

We present an extensive study of methods for exactly solving stochastic constraint (optimisation) problems (SCPs) in network analysis. These problems are prevalent in science, governance and industry. The first method we study is generic and decomposes stochastic constraints into a multitude of smaller local constraints that are solved using a constraint programming (CP) or mixed-integer programming (MIP) solver. However, many SCPs are formulated on probability distributions with a monotonic property, meaning that adding a positive decision to a partial solution to the problem cannot cause a decrease in solution quality. The second method is specifically designed for solving global stochastic constraints on monotonic probability distributions (SCMDs) in CP. Both methods use knowledge compilation to obtain a decision diagram encoding of the relevant probability distributions, where we focus on ordered binary decision diagrams (OBDDs). We discuss theoretical advantages and disadvantages of these methods and evaluate them experimentally. We observed that global approaches to solving SCMDs outperform decomposition approaches from CP, and perform complementarily to MIP-based decomposition approaches, while scaling much more favourably with instance size. Both methods have many alternative design choices, as both knowledge compilation and constraint solvers are used in a single pipeline. To identify which configurations work best, we apply programming by optimisation. Specifically, we show how an automated algorithm configurator can be used to find optimised configurations of our pipeline. After configuration, our global SCMD solving pipeline outperforms its closest competitor (a MIP-based decomposition pipeline) on all test sets we considered by up to two orders of magnitude in terms of PAR10 scores.

Bayesian optimization (BO) aims to minimize a given blackbox function using a model that is updated whenever new evidence about the function becomes available. Here, we address the problem of BO under partially right-censored response data, where in some evaluations we only obtain a lower bound on the function value. The ability to handle such response data allows us to adaptively censor costly function evaluations in minimization problems where the cost of a function evaluation corresponds to the function value. One important application giving rise to such censored data is the runtime-minimizing variant of the algorithm configuration problem: finding settings of a given parametric algorithm that minimize the runtime required for solving problem instances from a given distribution. We demonstrate that terminating slow algorithm runs prematurely and handling the resulting right-censored observations can substantially improve the state of the art in model-based algorithm configuration.

The performance of state-of-the-art algorithms is highly dependent on their parameter values, and choosing the right configuration can make the difference between solving a problem in a few minutes or hours. Automated algorithm configurators have shown their efficiency on a wide range of applications. However, they still encounter limitations when confronted to a large number of parameters to tune or long algorithm running time. We believe that there is untapped knowledge that can be gathered from the elements of the configuration problem, such as the default value in the configuration space, the source code of the algorithm, and the distribution of the problem instances at hand. We aim at utilising this knowledge to improve algorithm configurators.