In offline model-based optimisation (MBO) we are interested in using machine learning to design candidates that maximise some measure of desirability through an expensive but real-world scoring process. Offline MBO tries to approximate this expensive scoring function and use that to evaluate generated designs, however evaluation is non-exact because one approximation is being evaluated with another. Instead, we ask ourselves: if we did have the real world scoring function at hand, what cheap-to-compute validation metrics would correlate best with this? Since the real-world scoring function is available for simulated MBO datasets, insights obtained from this can be transferred over to real-world offline MBO tasks where the real-world scoring function is expensive to compute. To address this, we propose a conceptual evaluation framework that is amenable to measuring extrapolation, and apply this to conditional denoising diffusion models. Empirically, we find that two validation metrics -- agreement and Frechet distance -- correlate quite well with the ground truth. When there is high variability in conditional generation, feedback is required in the form of an approximated version of the real-world scoring function. Furthermore, we find that generating high-scoring samples may require heavily weighting the generative model in favour of sample quality, potentially at the cost of sample diversity.

We initiate the study of federated reinforcement learning under environmental heterogeneity by considering a policy evaluation problem. Our setup involves $N$ agents interacting with environments that share the same state and action space but differ in their reward functions and state transition kernels. Assuming agents can communicate via a central server, we ask: Does exchanging information expedite the process of evaluating a common policy? To answer this question, we provide the first comprehensive finite-time analysis of a federated temporal difference (TD) learning algorithm with linear function approximation, while accounting for Markovian sampling, heterogeneity in the agents' environments, and multiple local updates to save communication. Our analysis crucially relies on several novel ingredients: (i) deriving perturbation bounds on TD fixed points as a function of the heterogeneity in the agents' underlying Markov decision processes (MDPs); (ii) introducing a virtual MDP to closely approximate the dynamics of the federated TD algorithm; and (iii) using the virtual MDP to make explicit connections to federated optimization. Putting these pieces together, we rigorously prove that in a low-heterogeneity regime, exchanging model estimates leads to linear convergence speedups in the number of agents.

AI constructivism as inspired by Jean Piaget, described and surveyed by Frank Guerin, and representatively implemented by Gary Drescher seeks to create algorithms and knowledge structures that enable agents to acquire, maintain, and apply a deep understanding of the environment through sensorimotor interactions. This paper aims to increase awareness of constructivist AI implementations to encourage greater progress toward enabling lifelong learning by machines. It builds on Guerin's 2008 "Learning Like a Baby: A Survey of AI approaches." After briefly recapitulating that survey, it summarizes subsequent progress by the Guerin referents, numerous works not covered by Guerin (or found in other surveys), and relevant efforts in related areas. The focus is on knowledge representations and learning algorithms that have been used in practice viewed through lenses of Piaget's schemas, adaptation processes, and staged development. The paper concludes with a preview of a simple framework for constructive AI being developed by the author that parses concepts from sensory input and stores them in a semantic memory network linked to episodic data.

Quantifying similarity between neural representations -- e.g. hidden layer activation vectors -- is a perennial problem in deep learning and neuroscience research. Existing methods compare deterministic responses (e.g. artificial networks that lack stochastic layers) or averaged responses (e.g., trial-averaged firing rates in biological data). However, these measures of _deterministic_ representational similarity ignore the scale and geometric structure of noise, both of which play important roles in neural computation. To rectify this, we generalize previously proposed shape metrics (Williams et al. 2021) to quantify differences in _stochastic_ representations. These new distances satisfy the triangle inequality, and thus can be used as a rigorous basis for many supervised and unsupervised analyses. Leveraging this novel framework, we find that the stochastic geometries of neurobiological representations of oriented visual gratings and naturalistic scenes respectively resemble untrained and trained deep network representations. Further, we are able to more accurately predict certain network attributes (e.g. training hyperparameters) from its position in stochastic (versus deterministic) shape space.

To act in the world, robots rely on a representation of salient task aspects: for example, to carry a cup of coffee, a robot must consider movement efficiency and cup orientation in its behaviour. However, if we want robots to act for and with people, their representations must not be just functional but also reflective of what humans care about, i.e. their representations must be aligned with humans'. In this survey, we pose that current reward and imitation learning approaches suffer from representation misalignment, where the robot's learned representation does not capture the human's representation. We suggest that because humans will be the ultimate evaluator of robot performance in the world, it is critical that we explicitly focus our efforts on aligning learned task representations with humans, in addition to learning the downstream task. We advocate that current representation learning approaches in robotics should be studied from the perspective of how well they accomplish the objective of representation alignment. To do so, we mathematically define the problem, identify its key desiderata, and situate current robot learning methods within this formalism. We conclude the survey by suggesting future directions for exploring open challenges.

Language models trained on large text corpora encode rich distributional information about real-world environments and action sequences. This information plays a crucial role in current approaches to language processing tasks like question answering and instruction generation. We describe how to leverage language models for *non-linguistic* perception and control tasks. Our approach casts labeling and decision-making as inference in probabilistic graphical models in which language models parameterize prior distributions over labels, decisions and parameters, making it possible to integrate uncertain observations and incomplete background knowledge in a principled way. Applied to semantic segmentation, household navigation, and activity recognition tasks, this approach improves predictions on rare, out-of-distribution, and structurally novel inputs.

Despite the great interest in the bandit problem, designing efficient algorithms for complex models remains challenging, as there is typically no analytical way to quantify uncertainty. In this paper, we propose Multiplier Bootstrap-based Exploration (MBE), a novel exploration strategy that is applicable to any reward model amenable to weighted loss minimization. We prove both instance-dependent and instance-independent rate-optimal regret bounds for MBE in sub-Gaussian multi-armed bandits. With extensive simulation and real data experiments, we show the generality and adaptivity of MBE.

When we compare the influences of two causes on an outcome, if the conclusion from every group is against that from the conflation, we think there is Simpson's Paradox. The Existing Causal Inference Theory (ECIT) can make the overall conclusion consistent with the grouping conclusion by removing the confounder's influence to eliminate the paradox. The ECIT uses relative risk difference Pd = max(0, (R - 1)/R) (R denotes the risk ratio) as the probability of causation. In contrast, Philosopher Fitelson uses confirmation measure D (posterior probability minus prior probability) to measure the strength of causation. Fitelson concludes that from the perspective of Bayesian confirmation, we should directly accept the overall conclusion without considering the paradox. The author proposed a Bayesian confirmation measure b* similar to Pd before. To overcome the contradiction between the ECIT and Bayesian confirmation, the author uses the semantic information method with the minimum cross-entropy criterion to deduce causal confirmation measure Cc = (R -1)/max(R, 1). Cc is like Pd but has normalizing property (between -1 and 1) and cause symmetry. It especially fits cases where a cause restrains an outcome, such as the COVID-19 vaccine controlling the infection. Some examples (about kidney stone treatments and COVID-19) reveal that Pd and Cc are more reasonable than D; Cc is more useful than Pd.

In the stochastic multi-armed bandit problem, a randomized probability matching policy called Thompson sampling (TS) has shown excellent performance in various reward models. In addition to the empirical performance, TS has been shown to achieve asymptotic problem-dependent lower bounds in several models. However, its optimality has been mainly addressed under light-tailed or one-parameter models that belong to exponential families. In this paper, we consider the optimality of TS for the Pareto model that has a heavy tail and is parameterized by two unknown parameters. Specifically, we discuss the optimality of TS with probability matching priors that include the Jeffreys prior and the reference priors. We first prove that TS with certain probability matching priors can achieve the optimal regret bound. Then, we show the suboptimality of TS with other priors, including the Jeffreys and the reference priors. Nevertheless, we find that TS with the Jeffreys and reference priors can achieve the asymptotic lower bound if one uses a truncation procedure. These results suggest carefully choosing noninformative priors to avoid suboptimality and show the effectiveness of truncation procedures in TS-based policies.

We provide a theoretical justification for sample recovery using diffusion based image inpainting in a linear model setting. While most inpainting algorithms require retraining with each new mask, we prove that diffusion based inpainting generalizes well to unseen masks without retraining. We analyze a recently proposed popular diffusion based inpainting algorithm called RePaint (Lugmayr et al., 2022), and show that it has a bias due to misalignment that hampers sample recovery even in a two-state diffusion process. Motivated by our analysis, we propose a modified RePaint algorithm we call RePaint$^+$ that provably recovers the underlying true sample and enjoys a linear rate of convergence. It achieves this by rectifying the misalignment error present in drift and dispersion of the reverse process. To the best of our knowledge, this is the first linear convergence result for a diffusion based image inpainting algorithm.