Counterfactual learning from human bandit feedback describes a scenario where user feedback on the quality of outputs of a historic system is logged and used to improve a target system. We show how to apply this learning framework to neural semantic parsing. From a machine learning perspective, the key challenge lies in a proper reweighting of the estimator so as to avoid known degeneracies in counterfactual learning, while still being applicable to stochastic gradient optimization. To conduct experiments with human users, we devise an easy-to-use interface to collect human feedback on semantic parses. Our work is the first to show that semantic parsers can be improved significantly by counterfactual learning from logged human feedback data.

These successes evince an ability to accurately represent high dimensional functions, potentially of great use in computational and applied mathematics. That said, there are few rigorous results about the representation error and trainability of neural networks, as well as how they scale with the network size. Here we characterize both the error and scaling by reinterpreting the standard optimization algorithm used in machine learning applications, stochastic gradient descent, as the evolution of a particle system with interactions governed by a potential related to the objective or "loss" function used to train the network. We show that, when the number n of parameters is large, the empirical distribution of the particles descends on a convex landscape towards a minimizer at a rate independent of n . Our analysis also quantifies the scale and nature of the noise introduced by stochastic gradient descent and provides guidelines for the step size and batch size to use when training a neural network. We illustrate our findings on examples in which we train neural network to learn the energy function of the continuous 3-spin model on the sphere. The approximation error scales as our analysis predicts in as high a dimension as d 25. Finite training set and stochastic gradient descent 6 1.6. Central Limit Theorem (CL T) 13 3. Finite training set and stochastic gradient descent (SGD) 14 3.1. We thank Weinan E for discussions about the approximation error of neural networks and Sylvia Serfaty for her insights about interacting particle systems. M OTIVATION AND MAIN RESULTS While classification problems continue to be an active area research, extraordinary progress has been made on both speech and image recognition, problems that appeared intractable only a decade ago [1]. By harvesting the power of neural networks while simultaneously benefiting from advances in computational hardware, complex tasks such as automatic language translation are now routinely performed by computers with a high degree of reliability . The underlying explanation for these significant advances seems to be related to the expressive power of neural networks, and their ability to represent high dimensional functions with accuracy . These successes open exciting possibilities in applied and computational mathematics that are only beginning to be explored [2-8].

We introduce the SaaS Algorithm for semi-supervised learning, which uses learning speed during stochastic gradient descent in a deep neural network to measure the quality of an iterative estimate of the posterior probability of unknown labels. Training speed in supervised learning correlates strongly with the percentage of correct labels, so we use it as an inference criterion for the unknown labels, without attempting to infer the model parameters at first. Despite its simplicity, SaaS achieves state-of-the-art results in semi-supervised learning benchmarks.

Multi-agent settings are quickly gathering importance in machine learning. This includes a plethora of recent work on deep multi-agent reinforcement learning, but also can be extended to hierarchical RL, generative adversarial networks and decentralised optimisation. In all these settings the presence of multiple learning agents renders the training problem non-stationary and often leads to unstable training or undesired final results. We present Learning with Opponent-Learning Awareness (LOLA), a method in which each agent shapes the anticipated learning of the other agents in the environment. The LOLA learning rule includes an additional term that accounts for the impact of one agent's policy on the anticipated parameter update of the other agents. Preliminary results show that the encounter of two LOLA agents leads to the emergence of tit-for-tat and therefore cooperation in the iterated prisoners' dilemma, while independent learning does not. In this domain, LOLA also receives higher payouts compared to a naive learner, and is robust against exploitation by higher order gradient-based methods. Applied to repeated matching pennies, LOLA agents converge to the Nash equilibrium. In a round robin tournament we show that LOLA agents can successfully shape the learning of a range of multi-agent learning algorithms from literature, resulting in the highest average returns on the IPD. We also show that the LOLA update rule can be efficiently calculated using an extension of the policy gradient estimator, making the method suitable for model-free RL. This method thus scales to large parameter and input spaces and nonlinear function approximators. We also apply LOLA to a grid world task with an embedded social dilemma using deep recurrent policies and opponent modelling. Again, by explicitly considering the learning of the other agent, LOLA agents learn to cooperate out of self-interest.

In reinforcement learning, temporal difference (TD) is the most direct algorithm to learn the value function of a policy. For large or infinite state spaces, exact representations of the value function are usually not available, and it must be approximated by a function in some parametric family. However, with \emph{nonlinear} parametric approximations (such as neural networks), TD is not guaranteed to converge to a good approximation of the true value function within the family, and is known to diverge even in relatively simple cases. TD lacks an interpretation as a stochastic gradient descent of an error between the true and approximate value functions, which would provide such guarantees. We prove that approximate TD is a gradient descent provided the current policy is \emph{reversible}. This holds even with nonlinear approximations. A policy with transition probabilities $P(s,s')$ between states is reversible if there exists a function $\mu$ over states such that $\frac{P(s,s')}{P(s',s)}=\frac{\mu(s')}{\mu(s)}$. In particular, every move can be undone with some probability. This condition is restrictive; it is satisfied, for instance, for a navigation problem in any unoriented graph. In this case, approximate TD is exactly a gradient descent of the \emph{Dirichlet norm}, the norm of the difference of \emph{gradients} between the true and approximate value functions. The Dirichlet norm also controls the bias of approximate policy gradient. These results hold even with no decay factor ($\gamma=1$) and do not rely on contractivity of the Bellman operator, thus proving stability of TD even with $\gamma=1$ for reversible policies.

Matrix factorization is a well-studied task in machine learning for compactly representing large, noisy data. In our approach, instead of using the traditional concept of matrix rank, we define a new notion of link-rank based on a non-linear link function used within factorization. In particular, by applying the round function on a factorization to obtain ordinal-valued matrices, we introduce generalized round-rank (GRR). We show that not only are there many full-rank matrices that are low GRR, but further, that these matrices cannot be approximated well by low-rank linear factorization. We provide uniqueness conditions of this formulation and provide gradient descent-based algorithms. Finally, we present experiments on real-world datasets to demonstrate that the GRR-based factorization is significantly more accurate than linear factorization, while converging faster and using lower rank representations.

An evolution strategy (ES) variant based on a simplification of a natural evolution strategy recently attracted attention because it performs surprisingly well in challenging deep reinforcement learning domains. It searches for neural network parameters by generating perturbations to the current set of parameters, checking their performance, and moving in the aggregate direction of higher reward. Because it resembles a traditional finite-difference approximation of the reward gradient, it can naturally be confused with one. However, this ES optimizes for a different gradient than just reward: It optimizes for the average reward of the entire population, thereby seeking parameters that are robust to perturbation. This difference can channel ES into distinct areas of the search space relative to gradient descent, and also consequently to networks with distinct properties. This unique robustness-seeking property, and its consequences for optimization, are demonstrated in several domains. They include humanoid locomotion, where networks from policy gradient-based reinforcement learning are significantly less robust to parameter perturbation than ES-based policies solving the same task. While the implications of such robustness and robustness-seeking remain open to further study, this work's main contribution is to highlight such differences and their potential importance.

We consider securing a distributed machine learning system wherein the data is kept confidential by its providers who are recruited as workers to help the learner to train a $d$--dimensional model. In each communication round, up to $q$ out of the $m$ workers suffer Byzantine faults; faulty workers are assumed to have complete knowledge of the system and can collude to behave arbitrarily adversarially against the learner. We assume that each worker keeps a local sample of size $n$. (Thus, the total number of data points is $N=nm$.) Of particular interest is the high-dimensional regime $d \gg n$. We propose a secured variant of the classical gradient descent method which can tolerate up to a constant fraction of Byzantine workers. We show that the estimation error of the iterates converges to an estimation error $O(\sqrt{q/N} + \sqrt{d/N})$ in $O(\log N)$ rounds. The core of our method is a robust gradient aggregator based on the iterative filtering algorithm proposed by Steinhardt et al. \cite{Steinhardt18} for robust mean estimation. We establish a uniform concentration of the sample covariance matrix of gradients, and show that the aggregated gradient, as a function of model parameter, converges uniformly to the true gradient function. As a by-product, we develop a new concentration inequality for sample covariance matrices of sub-exponential distributions, which might be of independent interest.

We consider the problem of finding critical points of functions that are non-convex and non-smooth. Studying a fairly broad class of such problems, we analyze the behavior of three gradient-based methods (gradient descent, proximal update, and Frank-Wolfe update). For each of these methods, we establish rates of convergence for general problems, and also prove faster rates for continuous sub-analytic functions. We also show that our algorithms can escape strict saddle points for a class of non-smooth functions, thereby generalizing known results for smooth functions. Our analysis leads to a simplification of the popular CCCP algorithm, used for optimizing functions that can be written as a difference of two convex functions. Our simplified algorithm retains all the convergence properties of CCCP, along with a significantly lower cost per iteration. We illustrate our methods and theory via applications to the problems of best subset selection, robust estimation, mixture density estimation, and shape-from-shading reconstruction.

The learning algorithm of a neural network tries to optimize the neural network's weights until some stopping condition has been met. This condition is typically either when the error of the network reaches an acceptable level of accuracy on the training set, when the error of the network on the validation set begins to deteriorate, or when the specified computational budget has been exhausted. The most common learning algorithm for neural networks is back-propagation, an algorithm that uses stochastic gradient descent, which was discussed earlier on in this series. The are some problems with this approach. Adjusting all the weights at once can result in a significant movement of the neural network in weight space, the gradient descent algorithm is quite slow, and the gradient descent algorithm is susceptible to local minima.