Despite the considerable attention given to the questions of \textit{how much} and \textit{how to} explore in deep reinforcement learning, the investigation into \textit{when} to explore remains relatively less researched. While more sophisticated exploration strategies can excel in specific, often sparse reward environments, existing simpler approaches, such as $\epsilon$-greedy, persist in outperforming them across a broader spectrum of domains. The appeal of these simpler strategies lies in their ease of implementation and generality across a wide range of domains. The downside is that these methods are essentially a blind switching mechanism, which completely disregards the agent's internal state. In this paper, we propose to leverage the agent's internal state to decide \textit{when} to explore, addressing the shortcomings of blind switching mechanisms. We present Value Discrepancy and State Counts through homeostasis (VDSC), a novel approach for efficient exploration timing. Experimental results on the Atari suite demonstrate the superiority of our strategy over traditional methods such as $\epsilon$-greedy and Boltzmann, as well as more sophisticated techniques like Noisy Nets.
相關內容
I generalize state-of-the-art approaches that decompose differences in the distribution of a variable of interest between two groups into a portion explained by covariates and a residual portion. The method that I propose relaxes the overlapping supports assumption, allowing the groups being compared to not necessarily share exactly the same covariate support. I illustrate my method revisiting the black-white wealth gap in the U.S. as a function of labor income and other variables. Traditionally used decomposition methods would trim (or assign zero weight to) observations that lie outside the common covariate support region. On the other hand, by allowing all observations to contribute to the existing wealth gap, I find that otherwise trimmed observations contribute from 3% to 19% to the overall wealth gap, at different portions of the wealth distribution.
We give simply exponential lower bounds on the probabilities of a given strongly Rayleigh distribution, depending only on its expectation. This resolves a weak version of a problem left open by Karlin-Klein-Oveis Gharan in their recent breakthrough work on metric TSP, and this resolution leads to a minor improvement of their approximation factor for metric TSP. Our results also allow for a more streamlined analysis of the algorithm. To achieve these new bounds, we build upon the work of Gurvits-Leake on the use of the productization technique for bounding the capacity of a real stable polynomial. This technique allows one to reduce certain inequalities for real stable polynomials to products of affine linear forms, which have an underlying matrix structure. In this paper, we push this technique further by characterizing the worst-case polynomials via bipartitioned forests. This rigid combinatorial structure yields a clean induction argument, which implies our stronger bounds. In general, we believe the results of this paper will lead to further improvement and simplification of the analysis of various combinatorial and probabilistic bounds and algorithms.
An election is defined as a pair of a set of candidates C=\{c_1,\cdots,c_m\} and a multiset of votes V=\{v_1,\cdots,v_n\}, where each vote is a linear order of the candidates. The Borda election rule is characterized by a vector \langle m-1,m-2,\cdots,0\rangle, which means that the candidate ranked at the i-th position of a vote v receives a score m-i from v, and the candidate receiving the most score from all votes wins the election. Here, we consider the control problems of a Borda election, where the chair of the election attempts to influence the election outcome by adding or deleting either votes or candidates with the intention to make a special candidate win (constructive control) or lose (destructive control) the election. Control problems have been extensively studied for Borda elections from both classical and parameterized complexity viewpoints. We complete the parameterized complexity picture for Borda control problems by showing W[2]-hardness with the number of additions/deletions as parameter for constructive control by deleting votes, adding candidates, or deleting candidates. The hardness result for deleting votes settles an open problem posed by Liu and Zhu. Following the suggestion by Menon and Larson, we also investigate the impact of introducing top-truncated votes, where each voter ranks only t out of the given m candidates, on the classical and parameterized complexity of Borda control problems. Constructive Borda control problems remain NP-hard even with t being a small constant. Moreover, we prove that in the top-truncated case, constructive control by adding/deleting votes problems are FPT with the number \ell of additions/deletions and t as parameters, while for every constant t\geq 2, constructive control by adding/deleting candidates problems are W[2]-hard with respect to \ell.
Learning latent costs of transitions on graphs from trajectories demonstrations under various contextual features is challenging but useful for path planning. Yet, existing methods either oversimplify cost assumptions or scale poorly with the number of observed trajectories. This paper introduces DataSP, a differentiable all-to-all shortest path algorithm to facilitate learning latent costs from trajectories. It allows to learn from a large number of trajectories in each learning step without additional computation. Complex latent cost functions from contextual features can be represented in the algorithm through a neural network approximation. We further propose a method to sample paths from DataSP in order to reconstruct/mimic observed paths' distributions. We prove that the inferred distribution follows the maximum entropy principle. We show that DataSP outperforms state-of-the-art differentiable combinatorial solver and classical machine learning approaches in predicting paths on graphs.
Attention mechanism has been crucial for image diffusion models, however, their quadratic computational complexity limits the sizes of images we can process within reasonable time and memory constraints. This paper investigates the importance of dense attention in generative image models, which often contain redundant features, making them suitable for sparser attention mechanisms. We propose a novel training-free method ToDo that relies on token downsampling of key and value tokens to accelerate Stable Diffusion inference by up to 2x for common sizes and up to 4.5x or more for high resolutions like 2048x2048. We demonstrate that our approach outperforms previous methods in balancing efficient throughput and fidelity.
Graphs are used widely to model complex systems, and detecting anomalies in a graph is an important task in the analysis of complex systems. Graph anomalies are patterns in a graph that do not conform to normal patterns expected of the attributes and/or structures of the graph. In recent years, graph neural networks (GNNs) have been studied extensively and have successfully performed difficult machine learning tasks in node classification, link prediction, and graph classification thanks to the highly expressive capability via message passing in effectively learning graph representations. To solve the graph anomaly detection problem, GNN-based methods leverage information about the graph attributes (or features) and/or structures to learn to score anomalies appropriately. In this survey, we review the recent advances made in detecting graph anomalies using GNN models. Specifically, we summarize GNN-based methods according to the graph type (i.e., static and dynamic), the anomaly type (i.e., node, edge, subgraph, and whole graph), and the network architecture (e.g., graph autoencoder, graph convolutional network). To the best of our knowledge, this survey is the first comprehensive review of graph anomaly detection methods based on GNNs.
The existence of representative datasets is a prerequisite of many successful artificial intelligence and machine learning models. However, the subsequent application of these models often involves scenarios that are inadequately represented in the data used for training. The reasons for this are manifold and range from time and cost constraints to ethical considerations. As a consequence, the reliable use of these models, especially in safety-critical applications, is a huge challenge. Leveraging additional, already existing sources of knowledge is key to overcome the limitations of purely data-driven approaches, and eventually to increase the generalization capability of these models. Furthermore, predictions that conform with knowledge are crucial for making trustworthy and safe decisions even in underrepresented scenarios. This work provides an overview of existing techniques and methods in the literature that combine data-based models with existing knowledge. The identified approaches are structured according to the categories integration, extraction and conformity. Special attention is given to applications in the field of autonomous driving.
Deep learning has shown great potential for modeling the physical dynamics of complex particle systems such as fluids (in Lagrangian descriptions). Existing approaches, however, require the supervision of consecutive particle properties, including positions and velocities. In this paper, we consider a partially observable scenario known as fluid dynamics grounding, that is, inferring the state transitions and interactions within the fluid particle systems from sequential visual observations of the fluid surface. We propose a differentiable two-stage network named NeuroFluid. Our approach consists of (i) a particle-driven neural renderer, which involves fluid physical properties into the volume rendering function, and (ii) a particle transition model optimized to reduce the differences between the rendered and the observed images. NeuroFluid provides the first solution to unsupervised learning of particle-based fluid dynamics by training these two models jointly. It is shown to reasonably estimate the underlying physics of fluids with different initial shapes, viscosity, and densities. It is a potential alternative approach to understanding complex fluid mechanics, such as turbulence, that are difficult to model using traditional methods of mathematical physics.
In multi-turn dialog, utterances do not always take the full form of sentences \cite{Carbonell1983DiscoursePA}, which naturally makes understanding the dialog context more difficult. However, it is essential to fully grasp the dialog context to generate a reasonable response. Hence, in this paper, we propose to improve the response generation performance by examining the model's ability to answer a reading comprehension question, where the question is focused on the omitted information in the dialog. Enlightened by the multi-task learning scheme, we propose a joint framework that unifies these two tasks, sharing the same encoder to extract the common and task-invariant features with different decoders to learn task-specific features. To better fusing information from the question and the dialog history in the encoding part, we propose to augment the Transformer architecture with a memory updater, which is designed to selectively store and update the history dialog information so as to support downstream tasks. For the experiment, we employ human annotators to write and examine a large-scale dialog reading comprehension dataset. Extensive experiments are conducted on this dataset, and the results show that the proposed model brings substantial improvements over several strong baselines on both tasks. In this way, we demonstrate that reasoning can indeed help better response generation and vice versa. We release our large-scale dataset for further research.
With the capability of modeling bidirectional contexts, denoising autoencoding based pretraining like BERT achieves better performance than pretraining approaches based on autoregressive language modeling. However, relying on corrupting the input with masks, BERT neglects dependency between the masked positions and suffers from a pretrain-finetune discrepancy. In light of these pros and cons, we propose XLNet, a generalized autoregressive pretraining method that (1) enables learning bidirectional contexts by maximizing the expected likelihood over all permutations of the factorization order and (2) overcomes the limitations of BERT thanks to its autoregressive formulation. Furthermore, XLNet integrates ideas from Transformer-XL, the state-of-the-art autoregressive model, into pretraining. Empirically, XLNet outperforms BERT on 20 tasks, often by a large margin, and achieves state-of-the-art results on 18 tasks including question answering, natural language inference, sentiment analysis, and document ranking.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191