Reinforcement Learning (RL) enables an intelligent agent to optimise its performance in a task by continuously taking action from an observed state and receiving a feedback from the environment in form of rewards. RL typically uses tables or linear approximators to map state-action tuples that maximises the reward. Combining RL with deep neural networks (DRL) significantly increases its scalability and enables it to address more complex problems than before. However, DRL also inherits downsides from both RL and deep learning. Despite DRL improves generalisation across similar state-action pairs when compared to simpler RL policy representations like tabular methods, it still requires the agent to adequately explore the state-action space. Additionally, deep methods require more training data, with the volume of data escalating with the complexity and size of the neural network. As a result, deep RL requires a long time to collect enough agent-environment samples and to successfully learn the underlying policy. Furthermore, often even a slight alteration to the task invalidates any previous acquired knowledge. To address these shortcomings, Transfer Learning (TL) has been introduced, which enables the use of external knowledge from other tasks or agents to enhance a learning process. The goal of TL is to reduce the learning complexity for an agent dealing with an unfamiliar task by simplifying the exploration process. This is achieved by lowering the amount of new information required by its learning model, resulting in a reduced overall convergence time...

Active Learning (AL) is a sequential learning approach aiming at selecting the most informative data for model training. In many systems, safety constraints appear during data evaluation, requiring the development of safe AL methods. Key challenges of AL are the repeated model training and acquisition optimization required for data selection, which become particularly restrictive under safety constraints. This repeated effort often creates a bottleneck, especially in physical systems requiring real-time decision-making. In this paper, we propose a novel amortized safe AL framework. By leveraging a pretrained neural network policy, our method eliminates the need for repeated model training and acquisition optimization, achieving substantial speed improvements while maintaining competitive learning outcomes and safety awareness. The policy is trained entirely on synthetic data utilizing a novel safe AL objective. The resulting policy is highly versatile and adapts to a wide range of systems, as we demonstrate in our experiments. Furthermore, our framework is modular and we empirically show that we also achieve superior performance for unconstrained time-sensitive AL tasks if we omit the safety requirement.

Federated learning collaboratively trains a neural network on a global server, where each local client receives the current global model weights and sends back parameter updates (gradients) based on its local private data. The process of sending these model updates may leak client's private data information. Existing gradient inversion attacks can exploit this vulnerability to recover private training instances from a client's gradient vectors. Recently, researchers have proposed advanced gradient inversion techniques that existing defenses struggle to handle effectively. In this work, we present a novel defense tailored for large neural network models. Our defense capitalizes on the high dimensionality of the model parameters to perturb gradients within a subspace orthogonal to the original gradient. By leveraging cold posteriors over orthogonal subspaces, our defense implements a refined gradient update mechanism. This enables the selection of an optimal gradient that not only safeguards against gradient inversion attacks but also maintains model utility. We conduct comprehensive experiments across three different datasets and evaluate our defense against various state-of-the-art attacks and defenses. Code is available at https://censor-gradient.github.io.

Bayesian neural networks (BNNs) offer the potential for reliable uncertainty quantification and interpretability, which are critical for trustworthy AI in high-stakes domains. In this work, we advance message passing (MP) for BNNs and present a novel framework that models the predictive posterior as a factor graph. To the best of our knowledge, our framework is the first MP method that handles convolutional neural networks and avoids double-counting training data, a limitation of previous MP methods that causes overconfidence. We evaluate our approach on CIFAR-10 with a convolutional neural network of roughly 890k parameters and find that it can compete with the SOTA baselines AdamW and IVON, even having an edge in terms of calibration. On synthetic data, we validate the uncertainty estimates and observe a strong correlation (0.9) between posterior credible intervals and its probability of covering the true data-generating function outside the training range. While our method scales to an MLP with 5.6 million parameters, further improvements are necessary to match the scale and performance of state-of-the-art variational inference methods. Deep learning models have achieved impressive results across various domains, including natural language processing (Vaswani et al., 2023), computer vision (Ravi et al., 2024), and autonomous systems (Bojarski et al., 2016). Yet, they often produce overconfident but incorrect predictions, particularly in ambiguous or out-of-distribution scenarios. Without the ability to effectively quantify uncertainty, this can foster both overreliance and underreliance on models, as users stop trusting their outputs entirely (Zhang et al., 2024), and in high-stakes domains like healthcare or autonomous driving, its application can be dangerous (Henne et al., 2020). To ensure safer deployment in these settings, models must not only predict outcomes but also express how uncertain they are about those predictions to allow for informed decision-making. Bayesian neural networks (BNNs) offer a principled way to quantify uncertainty by capturing a posterior distribution over the model's weights, rather than relying on point estimates as in traditional neural networks. This allows BNNs to express epistemic uncertainty, the model's lack of knowledge about the underlying data distribution.

Exploration in sparse reward environments remains a significant challenge in reinforcement learning, particularly in Contextual Markov Decision Processes (CMDPs), where environments differ across episodes. Existing episodic intrinsic motivation methods for CMDPs primarily rely on count-based approaches, which are ineffective in large state spaces, or on similarity-based methods that lack appropriate metrics for state comparison. To address these shortcomings, we propose Episodic Novelty Through Temporal Distance (ETD), a novel approach that introduces temporal distance as a robust metric for state similarity and intrinsic reward computation. By employing contrastive learning, ETD accurately estimates temporal distances and derives intrinsic rewards based on the novelty of states within the current episode. Extensive experiments on various benchmark tasks demonstrate that ETD significantly outperforms state-of-the-art methods, highlighting its effectiveness in enhancing exploration in sparse reward CMDPs.

We introduce the problem of formally verifying properties of Markov processes where the parameters are the output of machine learning models. Our formulation is general and solves a wide range of problems, including verifying properties of probabilistic programs that use machine learning, and subgroup analysis in healthcare modeling. We show that for a broad class of machine learning models, including linear models, tree-based models, and neural networks, verifying properties of Markov chains like reachability, hitting time, and total reward can be formulated as a bilinear program. We develop a decomposition and bound propagation scheme for solving the bilinear program and show through computational experiments that our method solves the problem to global optimality up to 100x faster than state-of-the-art solvers. We also release $\texttt{markovml}$, an open-source tool for building Markov processes, integrating pretrained machine learning models, and verifying their properties, available at https://github.com/mmaaz-git/markovml.

Weaknesses: Overall, I thought this was a strong paper. The main concerns I had were as follows: (1) Mode-seeking versus showing the distribution: The aggregated results in the first experiment seem to show much more homogeneity than the results for GSP or MCMCP. It seems like one limitation of this approach might be that there is limited exploration of the space, perhaps making it hard to move between modes, and also makes it more difficult to see the full shape of the distribution, which I have often taken to be a goal in work using MCMCP. The movement between optimization and seeking a distribution is discussed to some extent in the paper, but I would be interested in seeing this discussed more (and perhaps whether GP without aggregation is likely to lead to more optimization than MCMCP). In the author response, they have shown additional information suggesting that GSP is more mode-seeking but also does a better job of capturing the distribution.

This paper introduces a new method for eliciting human representations of perceptual concepts, such as what RGB values people think correspond to the color "sunset" or what auditory dimensions (e.g. Rather than eliciting representations via guess-and-check (i.e., start with a dataset and then apply human-generated labels), this method (Gibbs Sampling with People, or GSP) enables inference to go in the other direction (i.e., start with labels, and then identify percepts that match those labels). GSP extends prior work (MCMC with People) to allow eliciting representations of much higher-dimensional stimuli. The reviewers unanimously praised this paper for tackling an important and relevant problem in cognitive science, for its breadth of empirical results, and for its novelty over prior work. R2 stated that the paper is "impressive in scale, scope, and results", R3 stated that it was "very relevant to the NeurIPS community and very novel", and R4 felt there could be "a potentially large impact of this work" with "substantial interest" amongst the NeurIPS community.

As the authors note, capacity is often balanced against robustness; since code redundancy is needed to enable recovery from noise, capacity is necessarily reduced because of the redundancy, making this an especially difficult problem. The authors rise to this challenge and claim to produce a network that exhibits "exponential capacity, robustness to large errors, and self decoding or clean-up of these errors". There network takes the structure of a restricted Boltzmann machine wherein the hidden units are comprised of clusters of neurons that laterally inhibit each other, each with the same connectivity to the input units but with different weights. The authors motivate their network using inspirations and ideas from expander graphs, error correcting codes, and Hopfield Networks. The proposed solution is straightforward and well-motivated, and all the design and algorithm choices seem quite sensible.

This paper presents a novel form of associative content addressable (ACA) memory systems. The canonical model for ACA memory is the Hopfield network, which can only store approximately N patterns of N bits. The authors use developments from error-correcting codes (ECCs) to implement an ACA that can store e N, N bit patterns. This is accomplished by using a bipartite expander graph, which is essentially a restricted Boltzmann machine (RBM) wherein the hidden nodes are actually clusters of units that are mutually inhibitory. The authors demonstrate that these networks have dynamics that can engage in error correction similar to ECCs, enabling the storage of exponentially many patterns.