The field of Automatic Machine Learning (AutoML) has recently attained impressive results, including the discovery of state-of-the-art machine learning solutions, such as neural image classifiers. This is often done by applying an evolutionary search method, which samples multiple candidate solutions from a large space and evaluates the quality of each candidate through a long training process. As a result, the search tends to be slow. In this paper, we show that large efficiency gains can be obtained by employing a fast unified functional hash, especially through the functional equivalence caching technique, which we also present. The central idea is to detect by hashing when the search method produces equivalent candidates, which occurs very frequently, and this way avoid their costly re-evaluation. Our hash is "functional" in that it identifies equivalent candidates even if they were represented or coded differently, and it is "unified" in that the same algorithm can hash arbitrary representations; e.g. compute graphs, imperative code, or lambda functions. As evidence, we show dramatic improvements on multiple AutoML domains, including neural architecture search and algorithm discovery. Finally, we consider the effect of hash collisions, evaluation noise, and search distribution through empirical analysis. Altogether, we hope this paper may serve as a guide to hashing techniques in AutoML.

Inductive programming frequently relies on some form of search in order to identify candidate solutions. However, the size of the search space limits the use of inductive programming to the production of relatively small programs. If we could somehow correctly predict the subset of instructions required for a given problem then inductive programming would be more tractable. We will show that this can be achieved in a high percentage of cases. This paper presents a novel model of programming language instruction co-occurrence that was built to support search space partitioning in the Zoea distributed inductive programming system. This consists of a collection of intersecting instruction subsets derived from a large sample of open source code. Using the approach different parts of the search space can be explored in parallel. The number of subsets required does not grow linearly with the quantity of code used to produce them and a manageable number of subsets is sufficient to cover a high percentage of unseen code. This approach also significantly reduces the overall size of the search space - often by many orders of magnitude.

We develop new tools in the theory of nonlinear random matrices and apply them to study the performance of the Sum of Squares (SoS) hierarchy on average-case problems. The SoS hierarchy is a powerful optimization technique that has achieved tremendous success for various problems in combinatorial optimization, robust statistics and machine learning. It's a family of convex relaxations that lets us smoothly trade off running time for approximation guarantees. In recent works, it's been shown to be extremely useful for recovering structure in high dimensional noisy data. It also remains our best approach towards refuting the notorious Unique Games Conjecture. In this work, we analyze the performance of the SoS hierarchy on fundamental problems stemming from statistics, theoretical computer science and statistical physics. In particular, we show subexponential-time SoS lower bounds for the problems of the Sherrington-Kirkpatrick Hamiltonian, Planted Slightly Denser Subgraph, Tensor Principal Components Analysis and Sparse Principal Components Analysis. These SoS lower bounds involve analyzing large random matrices, wherein lie our main contributions. These results offer strong evidence for the truth of and insight into the low-degree likelihood ratio hypothesis, an important conjecture that predicts the power of bounded-time algorithms for hypothesis testing. We also develop general-purpose tools for analyzing the behavior of random matrices which are functions of independent random variables. Towards this, we build on and generalize the matrix variant of the Efron-Stein inequalities. In particular, our general theorem on matrix concentration recovers various results that have appeared in the literature. We expect these random matrix theory ideas to have other significant applications.

Minimax optimization has seen a surge in interest with the advent of modern applications such as GANs, and it is inherently more challenging than simple minimization. The difficulty is exacerbated by the training data residing at multiple edge devices or \textit{clients}, especially when these clients can have heterogeneous datasets and local computation capabilities. We propose a general federated minimax optimization framework that subsumes such settings and several existing methods like Local SGDA. We show that naive aggregation of heterogeneous local progress results in optimizing a mismatched objective function -- a phenomenon previously observed in standard federated minimization. To fix this problem, we propose normalizing the client updates by the number of local steps undertaken between successive communication rounds. We analyze the convergence of the proposed algorithm for classes of nonconvex-concave and nonconvex-nonconcave functions and characterize the impact of heterogeneous client data, partial client participation, and heterogeneous local computations. Our analysis works under more general assumptions on the intra-client noise and inter-client heterogeneity than so far considered in the literature. For all the function classes considered, we significantly improve the existing computation and communication complexity results. Experimental results support our theoretical claims.

Why should moral philosophers, moral psychologists, and machine ethicists care about computational complexity? Debates on whether artificial intelligence (AI) can or should be used to solve problems in ethical domains have mainly been driven by what AI can or cannot do in terms of human capacities. In this paper, we tackle the problem from the other end by exploring what kind of moral machines are possible based on what computational systems can or cannot do. To do so, we analyze normative ethics through the lens of computational complexity. First, we introduce computational complexity for the uninitiated reader and discuss how the complexity of ethical problems can be framed within Marr's three levels of analysis. We then study a range of ethical problems based on consequentialism, deontology, and virtue ethics, with the aim of elucidating the complexity associated with the problems themselves (e.g., due to combinatorics, uncertainty, strategic dynamics), the computational methods employed (e.g., probability, logic, learning), and the available resources (e.g., time, knowledge, learning). The results indicate that most problems the normative frameworks pose lead to tractability issues in every category analyzed. Our investigation also provides several insights about the computational nature of normative ethics, including the differences between rule- and outcome-based moral strategies, and the implementation-variance with regard to moral resources. We then discuss the consequences complexity results have for the prospect of moral machines in virtue of the trade-off between optimality and efficiency. Finally, we elucidate how computational complexity can be used to inform both philosophical and cognitive-psychological research on human morality by advancing the Moral Tractability Thesis (MTT).

Inferring chemical reaction networks (CRN) from concentration time series is a challenge encouragedby the growing availability of quantitative temporal data at the cellular level. This motivates thedesign of algorithms to infer the preponderant reactions between the molecular species observed ina given biochemical process, and build CRN structure and kinetics models. Existing ODE-basedinference methods such as SINDy resort to least square regression combined with sparsity-enforcingpenalization, such as Lasso. However, we observe that these methods fail to learn sparse modelswhen the input time series are only available in wild type conditions, i.e. without the possibility toplay with combinations of zeroes in the initial conditions. We present a CRN inference algorithmwhich enforces sparsity by inferring reactions in a sequential fashion within a search tree of boundeddepth, ranking the inferred reaction candidates according to the variance of their kinetics on theirsupporting transitions, and re-optimizing the kinetic parameters of the CRN candidates on the wholetrace in a final pass. We show that Reactmine succeeds both on simulation data by retrievinghidden CRNs where SINDy fails, and on two real datasets, one of fluorescence videomicroscopyof cell cycle and circadian clock markers, the other one of biomedical measurements of systemiccircadian biomarkers possibly acting on clock gene expression in peripheral organs, by inferringpreponderant regulations in agreement with previous model-based analyses. The code is available athttps://gitlab.inria.fr/julmarti/crninf/ together with introductory notebooks.

In this paper, we extend the Descent framework, which enables learning and planning in the context of two-player games with perfect information, to the framework of stochastic games. We propose two ways of doing this, the first way generalizes the search algorithm, i.e. Descent, to stochastic games and the second way approximates stochastic games by deterministic games. We then evaluate them on the game EinStein wurfelt nicht! against state-of-the-art algorithms: Expectiminimax and Polygames (i.e. the Alpha Zero algorithm). It is our generalization of Descent which obtains the best results. The approximation by deterministic games nevertheless obtains good results, presaging that it could give better results in particular contexts.

Learning causal relationships between variables is a fundamental task in causal inference and directed acyclic graphs (DAGs) are a popular choice to represent the causal relationships. As one can recover a causal graph only up to its Markov equivalence class from observations, interventions are often used for the recovery task. Interventions are costly in general and it is important to design algorithms that minimize the number of interventions performed. In this work, we study the problem of identifying the smallest set of interventions required to learn the causal relationships between a subset of edges (target edges). Under the assumptions of faithfulness, causal sufficiency, and ideal interventions, we study this problem in two settings: when the underlying ground truth causal graph is known (subset verification) and when it is unknown (subset search). For the subset verification problem, we provide an efficient algorithm to compute a minimum sized interventional set; we further extend these results to bounded size non-atomic interventions and node-dependent interventional costs. For the subset search problem, in the worst case, we show that no algorithm (even with adaptivity or randomization) can achieve an approximation ratio that is asymptotically better than the vertex cover of the target edges when compared with the subset verification number. This result is surprising as there exists a logarithmic approximation algorithm for the search problem when we wish to recover the whole causal graph. To obtain our results, we prove several interesting structural properties of interventional causal graphs that we believe have applications beyond the subset verification/search problems studied here.

This work studies the combinatorial optimization problem of finding an optimal core tensor shape, also called multilinear rank, for a size-constrained Tucker decomposition. We give an algorithm with provable approximation guarantees for its reconstruction error via connections to higher-order singular values. Specifically, we introduce a novel Tucker packing problem, which we prove is NP-hard, and give a polynomial-time approximation scheme based on a reduction to the 2-dimensional knapsack problem with a matroid constraint. We also generalize our techniques to tree tensor network decompositions. We implement our algorithm using an integer programming solver, and show that its solution quality is competitive with (and sometimes better than) the greedy algorithm that uses the true Tucker decomposition loss at each step, while also running up to 1000x faster.

Word-level textual adversarial attacks have achieved striking performance in fooling natural language processing models. However, the fundamental questions of why these attacks are effective, and the intrinsic properties of the adversarial examples (AEs), are still not well understood. This work attempts to interpret textual attacks through the lens of $n$-gram frequency. Specifically, it is revealed that existing word-level attacks exhibit a strong tendency toward generation of examples with $n$-gram frequency descend ($n$-FD). Intuitively, this finding suggests a natural way to improve model robustness by training the model on the $n$-FD examples. To verify this idea, we devise a model-agnostic and gradient-free AE generation approach that relies solely on the $n$-gram frequency information, and further integrate it into the recently proposed convex hull framework for adversarial training. Surprisingly, the resultant method performs quite similarly to the original gradient-based method in terms of model robustness. These findings provide a human-understandable perspective for interpreting word-level textual adversarial attacks, and a new direction to improve model robustness.