This paper is about a new Bayesian method for multi label learning. The goal is to classify accurately in settings where there are many potential labels but only a few of them apply to each data point. The basis of the new results is a new generative model for the label vector of each example. Specifically the label vector y_n of the n-th example is generated as y_n f(V(\sigma(Wx_n)), where Wx_n is a lower dimensional projection of the n-th instance x_n, followed by an element-wise sigmoid activation \sigma. The final operation f corresponds to drawing Poisson random variables with rates given by V(\sigma(Wx_n)) and thresholding these so-called latent counts by taking the minimum with 1.

Weaknesses: - While the theoretical results seem correct, it is not clear to me the advantages of this approach compared to previous work, in particular, gradient Q-learning (GQ). On line 110, it is written that the assumptions are not as stringent but I am not convinced that this is the case. Could the authors clarify this point? If I am interpreting it correctly, it assumes that we have a fixed replay buffer of data on which we are doing updates, as in the offline batch RL setting. It is not specified which policy is used to collect this data and I would expect certain assumptions on this behavior policy.

This paper presents a new objective and an algorithm, which is similar to DQN, that optimises for that objective. Similar to prior work (GTD, TDC, GQ), the algorithm is shown to be convergent under linear function approximation. Because the objective is different, the paper could have better illustrated what this means in terms of the quality of the fixed point the new algorithm converges to - this is only discussed in detail in a special case of a diagonal feature covariance matrix. The author response did not lift this concern, and it remains unclear whether the new algorithm has major benefits over existing related work. The experiments were deemed somewhat insufficient to fully convince the reviewers of this.

Additional Feedback: After rebuttal: I read the author response and other reviews. It would be great to see more discussions in the next version. I think this paper has enough contribution and I will keep my original rating for acceptance. This paper is well written, and easy to follow. The main text is clean and the proof is deferred to the appendix.

All reviewers agree that the paper makes a nice contribution to planning with function approximation. In particular, the paper considers an important open problem, and while the problem is solved by making a few assumptions (mostly notably the core states), the results have made significant progress on the important problem. The reviewers also appreciate the use of precise language and careful description of related work. Among the remaining concerns, R2 wants to see some evidence of robustness against the failure of the "core state" assumption. While performing empirical experiments may not fit the theoretical nature of the paper, the authors can consider a theoretical justification: namely, define a notion of error that measures how much the core-states assumption is violated, and show how such an error manifest itself in the final guarantee.

This paper proposed the population posterior distribution for Bayesian modeling of streams of data and showed how stochastic optimization could be used to find a good approximation. The proposed framework and algorithm were demonstrated on both latent Dirichlet allocation and Dirichlet process mixture models on text and geolocation data and were shown to perform better than previous work in some cases. Overall, I think the main idea of the paper is very interesting and it would fit in well at NIPS. There are a few aspects of the paper that could use some more discussion though. First, the authors were very careful throughout the paper to use the term "Bayesian modeling", except the title uses "Bayesian inference", which this paper definitely does not provide a method for.

This paper extends previous work on modeling network interactions with multivariate Hawkes processes by introducing a time-varying network into the model. For example, when one Twitter user publishes a tweet, followers of that user are likely to retweet in response. The dynamic network is intended to capture the creation of new connections, for example, when one Twitter user begins to follow another. The instantaneous network is represented by a binary adjacency matrix, and edges are added to the network according to a survival process with a event-driven rate. If one user frequently retweets another user's messages, then it is likely they will begin to follow that user and thereby add a new connection to the network.

Artificial intelligence commonly refers to the science and engineering of artificial systems that can carry out tasks generally associated with requiring aspects of human intelligence, such as playing games, translating languages, and driving cars. In recent years, there have been exciting advances in learning-based, data-driven approaches towards AI, and machine learning and deep learning have enabled computer systems to perceive the world in unprecedented ways. Reinforcement learning has enabled breakthroughs in complex games such as Go and challenging robotics tasks such as quadrupedal locomotion. A key aspect of intelligence is to not only make predictions, but reason about the uncertainty in these predictions, and to consider this uncertainty when making decisions. This is what this manuscript on "Probabilistic Artificial Intelligence" is about. The first part covers probabilistic approaches to machine learning. We discuss the differentiation between "epistemic" uncertainty due to lack of data and "aleatoric" uncertainty, which is irreducible and stems, e.g., from noisy observations and outcomes. We discuss concrete approaches towards probabilistic inference and modern approaches to efficient approximate inference. The second part of the manuscript is about taking uncertainty into account in sequential decision tasks. We consider active learning and Bayesian optimization -- approaches that collect data by proposing experiments that are informative for reducing the epistemic uncertainty. We then consider reinforcement learning and modern deep RL approaches that use neural network function approximation. We close by discussing modern approaches in model-based RL, which harness epistemic and aleatoric uncertainty to guide exploration, while also reasoning about safety.

The Shapley value is the prevalent solution for fair division problems in which a payout is to be divided among multiple agents. By adopting a game-theoretic view, the idea of fair division and the Shapley value can also be used in machine learning to quantify the individual contribution of features or data points to the performance of a predictive model. Despite its popularity and axiomatic justification, the Shapley value suffers from a computational complexity that scales exponentially with the number of entities involved, and hence requires approximation methods for its reliable estimation. We propose SVA$k_{\text{ADD}}$, a novel approximation method that fits a $k$-additive surrogate game. By taking advantage of $k$-additivity, we are able to elicit the exact Shapley values of the surrogate game and then use these values as estimates for the original fair division problem. The efficacy of our method is evaluated empirically and compared to competing methods.

We introduce in-context denoising, a task that refines the connection between attention-based architectures and dense associative memory (DAM) networks, also known as modern Hopfield networks. Using a Bayesian framework, we show theoretically and empirically that certain restricted denoising problems can be solved optimally even by a single-layer transformer. We demonstrate that a trained attention layer processes each denoising prompt by performing a single gradient descent update on a context-aware DAM energy landscape, where context tokens serve as associative memories and the query token acts as an initial state. This one-step update yields better solutions than exact retrieval of either a context token or a spurious local minimum, providing a concrete example of DAM networks extending beyond the standard retrieval paradigm. Overall, this work solidifies the link between associative memory and attention mechanisms first identified by Ramsauer et al., and demonstrates the relevance of associative memory models in the study of in-context learning.