Tuning machine learning hyperparameters is a tedious yet crucial task, as the performance of an algorithm can be highly dependent on the choice of hyperparameters. Manual tuning takes time away from important steps of the machine learning pipeline like feature engineering and interpreting results. Grid and random search are hands-off, but require long run times because they waste time evaluating unpromising areas of the search space. Increasingly, hyperparameter tuning is done by automated methods that aim to find optimal hyperparameters in less time using an informed search with no manual effort necessary beyond the initial set-up. Bayesian optimization, a model-based method for finding the minimum of a function, has recently been applied to machine learning hyper parameter tuning, with results suggesting this approach can achieve better performance on the test set while requiring fewer iterations than random search.

With the rapid rise of neural architecture search, the ability to understand its complexity from the perspective of a search algorithm is desirable. Recently, Traor\'e et al. have proposed the framework of Fitness Landscape Footprint to help describe and compare neural architecture search problems. It attempts at describing why a search strategy might be successful, struggle or fail on a target task. Our study leverages this methodology in the context of searching across sensors, including sensor data fusion. In particular, we apply the Fitness Landscape Footprint to the real-world image classification problem of So2Sat LCZ42, in order to identify the most beneficial sensor to our neural network hyper-parameter optimization problem. From the perspective of distributions of fitness, our findings indicate a similar behaviour of the search space for all sensors: the longer the training time, the larger the overall fitness, and more flatness in the landscapes (less ruggedness and deviation). Regarding sensors, the better the fitness they enable (Sentinel-2), the better the search trajectories (smoother, higher persistence). Results also indicate very similar search behaviour for sensors that can be decently fitted by the search space (Sentinel-2 and fusion).

This manuscript introduces an autotuned algorithm for searching nearest neighbors based on neighbor graphs and optimization metaheuristics to produce Pareto-optimal searches for quality and search speed automatically; the same strategy is also used to produce indexes that achieve a minimum quality. Our approach is described and benchmarked with other state-of-the-art similarity search methods, showing convenience and competitiveness.

Educational Timetabling, in essence, consists in assigning teacher/student meetings to days, timeslots, and classrooms. Despite this apparent simplicity, experience teaches us that every single institution has its own rules, conventions, and fixations, thus making each specific problem almost unique. As a consequence, uncountably many different problem formulations have been proposed in the literature on Educational Timetabling, depending on the type of institution (high-school, university, or other), the type of meetings (lectures, exams,...), and the different settings, constraints, and objectives. Many papers in the literature tackle a specific problem using a selected search method. The authors normally claim the success of the application, though rarely dispelling the doubt over the readers that the method used was more the authors' "favorite" rather than the most suitable for the problem under consideration.

We study the bilinearly coupled minimax problem: $\min_{x} \max_{y} f(x) + y^\top A x - h(y)$, where $f$ and $h$ are both strongly convex smooth functions and admit first-order gradient oracles. Surprisingly, no known first-order algorithms have hitherto achieved the lower complexity bound of $\Omega((\sqrt{\frac{L_x}{\mu_x}} + \frac{\|A\|}{\sqrt{\mu_x \mu_y}} + \sqrt{\frac{L_y}{\mu_y}}) \log(\frac1{\varepsilon}))$ for solving this problem up to an $\varepsilon$ primal-dual gap in the general parameter regime, where $L_x, L_y,\mu_x,\mu_y$ are the corresponding smoothness and strongly convexity constants. We close this gap by devising the first optimal algorithm, the Lifted Primal-Dual (LPD) method. Our method lifts the objective into an extended form that allows both the smooth terms and the bilinear term to be handled optimally and seamlessly with the same primal-dual framework. Besides optimality, our method yields a desirably simple single-loop algorithm that uses only one gradient oracle call per iteration. Moreover, when $f$ is just convex, the same algorithm applied to a smoothed objective achieves the nearly optimal iteration complexity. We also provide a direct single-loop algorithm, using the LPD method, that achieves the iteration complexity of $O(\sqrt{\frac{L_x}{\varepsilon}} + \frac{\|A\|}{\sqrt{\mu_y \varepsilon}} + \sqrt{\frac{L_y}{\varepsilon}})$. Numerical experiments on quadratic minimax problems and policy evaluation problems further demonstrate the fast convergence of our algorithm in practice.

Finding an optimal set of critical nodes in a complex network has been a long-standing problem in the fields of both artificial intelligence and operations research. Potential applications include epidemic control, network security, carbon emission monitoring, emergence response, drug design, and vulnerability assessment. In this work, we consider the problem of finding a minimal node separator whose removal separates a graph into multiple different connected components with fewer than a limited number of vertices in each component. To solve it, we propose a frequent itemset-driven search approach, which integrates the concept of frequent itemset mining in data mining into the well-known memetic search framework. Starting from a high-quality population built by the solution construction and population repair procedures, it iteratively employs the frequent itemset recombination operator (to generate promising offspring solution based on itemsets that frequently occur in high-quality solutions), tabu search-based simulated annealing (to find high-quality local optima), population repair procedure (to modify the population), and rank-based population management strategy (to guarantee a healthy population). Extensive evaluations on 50 widely used benchmark instances show that it significantly outperforms state-of-the-art algorithms. In particular, it discovers 29 new upper bounds and matches 18 previous best-known bounds. Finally, experimental analyses are performed to confirm the effectiveness of key algorithmic modules of the proposed method.

Reinforcement learning policies are often represented by neural networks, but programmatic policies are preferred in some cases because they are more interpretable, amenable to formal verification, or generalize better. While efficient algorithms for learning neural policies exist, learning programmatic policies is challenging. Combining imitation-projection and dataset aggregation with a local search heuristic, we present a simple and direct approach to extracting a programmatic policy from a pretrained neural policy. After examining our local search heuristic on a programming by example problem, we demonstrate our programmatic policy extraction method on a pendulum swing-up problem. Both when trained using a hand crafted expert policy and a learned neural policy, our method discovers simple and interpretable policies that perform almost as well as the original.

Arguably the most fundamental question in the theory of generative adversarial networks (GANs) is to understand to what extent GANs can actually learn the underlying distribution. Theoretical and empirical evidence suggests local optimality of the empirical training objective is insufficient. Yet, it does not rule out the possibility that achieving a true population minimax optimal solution might imply distribution learning. In this paper, we show that standard cryptographic assumptions imply that this stronger condition is still insufficient. Namely, we show that if local pseudorandom generators (PRGs) exist, then for a large family of natural continuous target distributions, there are ReLU network generators of constant depth and polynomial size which take Gaussian random seeds so that (i) the output is far in Wasserstein distance from the target distribution, but (ii) no polynomially large Lipschitz discriminator ReLU network can detect this. This implies that even achieving a population minimax optimal solution to the Wasserstein GAN objective is likely insufficient for distribution learning in the usual statistical sense. Our techniques reveal a deep connection between GANs and PRGs, which we believe will lead to further insights into the computational landscape of GANs.

How resources are deployed to secure critical targets in networks can be modelled by Network Security Games (NSGs). While recent advances in deep learning (DL) provide a powerful approach to dealing with large-scale NSGs, DL methods such as NSG-NFSP suffer from the problem of data inefficiency. Furthermore, due to centralized control, they cannot scale to scenarios with a large number of resources. In this paper, we propose a novel DL-based method, NSGZero, to learn a non-exploitable policy in NSGs. NSGZero improves data efficiency by performing planning with neural Monte Carlo Tree Search (MCTS). Our main contributions are threefold. First, we design deep neural networks (DNNs) to perform neural MCTS in NSGs. Second, we enable neural MCTS with decentralized control, making NSGZero applicable to NSGs with many resources. Third, we provide an efficient learning paradigm, to achieve joint training of the DNNs in NSGZero. Compared to state-of-the-art algorithms, our method achieves significantly better data efficiency and scalability.

Online advertising is a major source of income for many online companies. One common approach is to sell online advertisements via waterfall auctions, where a publisher makes sequential price offers to ad networks. The publisher controls the order and prices of the waterfall and by that aims to maximize his revenue. In this work, we propose a methodology to learn a waterfall strategy from historical data by wisely searching in the space of possible waterfalls and selecting the one leading to the highest revenue. The contribution of this work is twofold; First, we propose a novel method to estimate the valuation distribution of each user with respect to each ad network. Second, we utilize the valuation matrix to score our candidate waterfalls as part of a procedure that iteratively searches in local neighborhoods. Our framework guarantees that the waterfall revenue improves between iterations until converging to a local optimum. Real-world demonstrations are provided to show that the proposed method improves the total revenue of real-world waterfalls compared to manual expert optimization. Finally, the code and the data are available here.