Recent developments in sequential experimental design look to construct a policy that can efficiently navigate the design space, in a way that maximises the expected information gain. Whilst there is work on achieving tractable policies for experimental design problems, there is significantly less work on obtaining policies that are able to generalise well - i.e. able to give good performance despite a change in the underlying statistical properties of the experiments. Conducting experiments sequentially has recently brought about the use of reinforcement learning, where an agent is trained to navigate the design space to select the most informative designs for experimentation. However, there is still a lack of understanding about the benefits and drawbacks of using certain reinforcement learning algorithms to train these agents. In our work, we investigate several reinforcement learning algorithms and their efficacy in producing agents that take maximally informative design decisions in sequential experimental design scenarios. We find that agent performance is impacted depending on the algorithm used for training, and that particular algorithms, using dropout or ensemble approaches, empirically showcase attractive generalisation properties.

Individual treatment effect (ITE) estimation is to evaluate the causal effects of treatment strategies on some important outcomes, which is a crucial problem in healthcare. Most existing ITE estimation methods are designed for centralized settings. However, in real-world clinical scenarios, the raw data are usually not shareable among hospitals due to the potential privacy and security risks, which makes the methods not applicable. In this work, we study the ITE estimation task in a federated setting, which allows us to harness the decentralized data from multiple hospitals. Due to the unavoidable confounding bias in the collected data, a model directly learned from it would be inaccurate. One well-known solution is Inverse Probability Treatment Weighting (IPTW), which uses the conditional probability of treatment given the covariates to re-weight each training example. Applying IPTW in a federated setting, however, is non-trivial. We found that even with a well-estimated conditional probability, the local model training step using each hospital's data alone would still suffer from confounding bias. To address this, we propose FED-IPTW, a novel algorithm to extend IPTW into a federated setting that enforces both global (over all the data) and local (within each hospital) decorrelation between covariates and treatments. We validated our approach on the task of comparing the treatment effects of mechanical ventilation on improving survival probability for patients with breadth difficulties in the intensive care unit (ICU). We conducted experiments on both synthetic and real-world eICU datasets and the results show that FED-IPTW outperform state-of-the-art methods on all the metrics on factual prediction and ITE estimation tasks, paving the way for personalized treatment strategy design in mechanical ventilation usage.

Preprint submitted to Elsevier March 10, 2025 algorithms to solve such inverse-type problems advance different fields including inferring neural circuit dynamics from spiking data [42] in neuroscience, modeling and predicting complex weather patterns from historical data [9] in climate science, uncovering disease transmission dynamics from infection case counts over time [46] in epidemiology, and deducing reaction rates from experimental concentration-time profiles in reaction kinetics in biochemistry [30]. However, such inverse-type problems pose substantial mathematical and computational challenges, particularly when data are limited and noisy, motivating ongoing research into novel algorithms and theoretical frameworks to improve models' reconstruction accuracy and efficiency. In this paper, we study the inverse problem of inferring the distribution of model parameters for several dynamical systems including ordinary differential equations (ODEs), partial differential equations (PDEs), and stochastic differential equations (SDEs) from time-series data or spatiotemporal data. Existing methods for such problems can be broadly categorized into traditional statistical approaches and modern data-driven techniques. Traditional statistical methods often involve parameter estimation frameworks. For example, linear and nonlinear regression methods play a role in simpler systems where the functional form of the model is partially known [13]. Furthermore, maximum likelihood estimation and Bayesian inference methods [16, 33] are often adopted. Maximum likelihood estimation optimizes the likelihood of model parameter values in a proposed model from observed data, while Bayesian methods incorporate prior information and compute posterior distributions. These approaches are widely used in applications such as reaction network reconstruction and epidemiological modeling.

Large language models (LLMs) exhibit the ability to generalize given few-shot examples in their input prompt, an emergent capability known as in-context learning (ICL). We investigate whether LLMs utilize ICL to perform structured reasoning in ways that are consistent with a Bayesian framework or rely on pattern matching. Using a controlled setting of biased coin flips, we find that: (1) LLMs often possess biased priors, causing initial divergence in zero-shot settings, (2) in-context evidence outweighs explicit bias instructions, (3) LLMs broadly follow Bayesian posterior updates, with deviations primarily due to miscalibrated priors rather than flawed updates, and (4) attention magnitude has negligible effect on Bayesian inference. With sufficient demonstrations of biased coin flips via ICL, LLMs update their priors in a Bayesian manner.

Machine learning systems struggle with robustness, under subpopulation shifts. This problem becomes especially pronounced in scenarios where only a subset of attribute combinations is observed during training -a severe form of subpopulation shift, referred as compositional shift. To address this problem, we ask the following question: Can we improve the robustness by training on synthetic data, spanning all possible attribute combinations? We first show that training of conditional diffusion models on limited data lead to incorrect underlying distribution. Therefore, synthetic data sampled from such models will result in unfaithful samples and does not lead to improve performance of downstream machine learning systems. To address this problem, we propose CoInD to reflect the compositional nature of the world by enforcing conditional independence through minimizing Fisher's divergence between joint and marginal distributions. We demonstrate that synthetic data generated by CoInD is faithful and this translates to state-of-the-art worst-group accuracy on compositional shift tasks on CelebA.

With the rapid development of embodied intelligence, locomotion control of quadruped robots on complex terrains has become a research hotspot. Unlike traditional locomotion control approaches focusing solely on velocity tracking, we pursue to balance the agility and robustness of quadruped robots on diverse and complex terrains. To this end, we propose an end-to-end deep reinforcement learning framework for posture-aware locomotion named PALo, which manages to handle simultaneous linear and angular velocity tracking and real-time adjustments of body height, pitch, and roll angles. In PALo, the locomotion control problem is formulated as a partially observable Markov decision process, and an asymmetric actor-critic architecture is adopted to overcome the sim-to-real challenge. Further, by incorporating customized training curricula, PALo achieves agile posture-aware locomotion control in simulated environments and successfully transfers to real-world settings without fine-tuning, allowing real-time control of the quadruped robot's locomotion and body posture across challenging terrains. Through in-depth experimental analysis, we identify the key components of PALo that contribute to its performance, further validating the effectiveness of the proposed method. The results of this study provide new possibilities for the low-level locomotion control of quadruped robots in higher dimensional command spaces and lay the foundation for future research on upper-level modules for embodied intelligence.

Concept-based models are an emerging paradigm in deep learning that constrains the inference process to operate through human-interpretable concepts, facilitating explainability and human interaction. However, these architectures, on par with popular opaque neural models, fail to account for the true causal mechanisms underlying the target phenomena represented in the data. This hampers their ability to support causal reasoning tasks, limits out-of-distribution generalization, and hinders the implementation of fairness constraints. To overcome these issues, we propose \emph{Causally reliable Concept Bottleneck Models} (C$^2$BMs), a class of concept-based architectures that enforce reasoning through a bottleneck of concepts structured according to a model of the real-world causal mechanisms. We also introduce a pipeline to automatically learn this structure from observational data and \emph{unstructured} background knowledge (e.g., scientific literature). Experimental evidence suggest that C$^2$BM are more interpretable, causally reliable, and improve responsiveness to interventions w.r.t. standard opaque and concept-based models, while maintaining their accuracy.

Reinforcement Learning (RL) and Multi-Agent Reinforcement Learning (MARL) offer promising solutions for complex, dynamic environments where decision-making entities must interact and adapt. In cybersecurity, particularly in Automated Cyber Defence(ACD), these techniques can address challenges posed by high-dimensional observations and actions. This document provides guidelines for: Cybersecurity professionals exploring RL and MARL for real-world applications. RL and MARL researchers aiming to tackle the nuanced demands of cybersecurity scenarios. By outlining when RL and MARL are appropriate, addressing cyber-specific challenges, and offering practical considerations for implementation, these guidelines aim to bridge the gap between theoretical research and practical deployment in adversarial settings. We expect that this document will offer support to researchers who are keen to explore topics at the intersection of RL, MARL and ACD by highlighting open research questions and topics that demand further investigation.

The complexity of online decision-making under uncertainty stems from the requirement of finding a balance between exploiting known strategies and exploring new possibilities. Naturally, the uncertainty type plays a crucial role in developing decision-making strategies that manage complexity effectively. In this paper, we focus on a specific form of uncertainty known as epistemic ambivalence (EA), which emerges from conflicting pieces of evidence or contradictory experiences. It creates a delicate interplay between uncertainty and confidence, distinguishing it from epistemic uncertainty that typically diminishes with new information. Indeed, ambivalence can persist even after additional knowledge is acquired. To address this phenomenon, we propose a novel framework, called the epistemically ambivalent Markov decision process (EA-MDP), aiming to understand and control EA in decision-making processes. This framework incorporates the concept of a quantum state from the quantum mechanics formalism, and its core is to assess the probability and reward of every possible outcome. We calculate the reward function using quantum measurement techniques and prove the existence of an optimal policy and an optimal value function in the EA-MDP framework. We also propose the EA-epsilon-greedy Q-learning algorithm. To evaluate the impact of EA on decision-making and the expedience of our framework, we study two distinct experimental setups, namely the two-state problem and the lattice problem. Our results show that using our methods, the agent converges to the optimal policy in the presence of EA.

We extend a recently introduced geometric interpretation of Markov Decision Processes (MDPs) that provides a new perspective on MDP algorithms and their dynamics. Based on this view, we develop a novel analytical framework that simplifies the proofs of existing results and enables us to derive new ones. Specifically, we analyze the behavior of two classical MDP-solving algorithms: Policy Iteration (PI) and Value Iteration (VI). For each algorithm, we first describe its dynamics in geometric terms and then present an analysis along with several convergence results. We begin by introducing an MDP transformation that modifies the discount factor γ and demonstrate how this transformation improves the convergence properties of both algorithms, provided that it can be applied such that the resulting system remains a regular MDP. Second, we present a new analysis of PI in a 2-state MDP case, showing that the number of iterations required for convergence is bounded by the number of state-action pairs. Finally, we reveal an additional convergence factor in the VI algorithm for cases with a connected optimal policy, which is attributed to an extra rotation component in the VI dynamics.