Learning disentanglement aims at finding a low dimensional representation which consists of multiple explanatory and generative factors the observational data. The framework variational autoencoder (VAE) is commonly used to disentangle independent from observations. However, in real scenarios, with semantics are not necessarily independent. Instead, there might be an underlying causal structure renders these dependent. We thus propose new VAE based named CausalVAE, includes Causal Layer...

Top-N recommendation, which aims to learn user ranking-based preference, has long been a fundamental problem in wide range of applications. Traditional models usually motivate themselves by designing complex or tailored architectures based on different assumptions. However, the training data recommender system can be extremely sparse and imbalanced, poses great challenges for boosting recommendation performance. To alleviate this problem, paper, we propose reformulate task within causal...

In many real-world scenarios, such as gas leak detection or environmental pollutant tracking, solving the Inverse Source Localization and Characterization problem involves navigating complex, dynamic fields with sparse noisy observations. Traditional methods face significant challenges, including partial observability, temporal spatial dynamics, out-of-distribution generalization, reward sparsity. To address these issues, we propose a hierarchical framework that integrates Bayesian inference...

Particle filtering is a Bayesian inference method and fundamental tool in state estimation for dynamic systems, but its effectiveness often limited by the constraints of initial prior distribution, phenomenon we define as Prior Boundary Phenomenon. This challenge arises when target states lie outside prior's support, rendering traditional particle methods inadequate accurate estimation. Although techniques like unbounded priors larger sets have been proposed, they remain computationally...

In this article, according to the real-time and accuracy requirements of advanced vehicle-assisted driving in pedestrian detection, an improved LeNet-5 convolutional neural network is proposed. Firstly, structure model analyzed, parameters are optimized on basis get a new LeNet model, then it used detect pedestrians. Finally, miss rate found be 25% by contrast analysis. The experiment proves that method better than SA-Fast R-CNN classical CNN algorithm.

Accurately estimating the state of charge (SOC) power batteries in electric vehicles is great significance to measurement endurance mileage vehicles, as well safety protection battery. In view lithium ion batteries’ nonlinear relation between SOC estimation and current, voltage, temperature, improved Back Propagation (BP) neural network method proposed accurately estimate batteries. To address inherent limitations BP network, particle swarm algorithm adopted modify relevant weighting...

Debiased recommendation has recently attracted increasing attention from both industry and academic communities. Traditional models mostly rely on the inverse propensity score (IPS), which can be hard to estimate may suffer high variance issue. To alleviate these problems, in this article, we propose a novel debiased framework based user feature balancing. The general idea is introduce projection function adjust distributions, such that ideal unbiased learning objective upper bounded by...

Learning disentanglement aims at finding a low dimensional representation which consists of multiple explanatory and generative factors the observational data. The framework variational autoencoder (VAE) is commonly used to disentangle independent from observations. However, in real scenarios, with semantics are not necessarily independent. Instead, there might be an underlying causal structure renders these dependent. We thus propose new VAE based named CausalVAE, includes Causal Layer...

The issue of fairness in recommendation systems has recently become a matter growing concern for both the academic and industrial sectors due to potential bias machine learning models. One such is that feedback loops, where collection data from an unfair online system hinders accurate evaluation relevance scores between users items. Given often recommend popular content vendors, underlying items may not be accurately represented training data. Hence, this creates loop which user longer...

Contextual multi-armed bandit (MAB) achieves cutting-edge performance on a variety of problems. When it comes to real-world scenarios such as recommendation system and online advertising, however, is essential consider the resource consumption exploration. In practice, there typically non-zero cost associated with executing (arm) in environment, hence, policy should be learned fixed exploration constraint. It challenging learn global optimal directly, since NP-hard problem significantly...

Ranking items regarding individual user interests is a core technique of multiple downstream tasks such as recommender systems. Learning personalized ranker typically relies on the implicit feedback from users' past click-through behaviors. However, collected biased toward previously highly-ranked and directly learning it would result in "rich-get-richer" phenomena. In this paper, we propose simple yet sufficient unbiased learning-to-rank paradigm named InfoRank that aims to simultaneously...

Ranking items regarding individual user interests is a core technique of multiple downstream tasks such as recommender systems. Learning personalized ranker typically relies on the implicit feedback from users' past click-through behaviors. However, collected biased toward previously highly-ranked and directly learning it would result in "rich-get-richer" phenomenon. In this paper, we propose simple yet sufficient unbiased learning-to-rank paradigm named InfoRank that aims to simultaneously...

In human-AI interaction, a prominent goal is to attain human`s desirable outcome with the assistance of AI agents, which can be ideally delineated as problem seeking optimal Nash Equilibrium that matches outcome. However, reaching usually challenging due existence multiple Equilibria are related assisting task but do not correspond To tackle this issue, we employ theoretical framework called structural causal game (SCG) formalize interactive process. Furthermore, introduce strategy referred...

Learning representations with a high Probability of Necessary and Sufficient Causes (PNS) has been shown to enhance deep learning models' ability. This task involves identifying causal features that are both sufficient (guaranteeing the outcome) necessary (without which outcome cannot occur). However, current research predominantly focuses on unimodal data, extending PNS multimodal settings presents significant challenges. The challenges arise as conditions for identifiability, Exogeneity...

In sequential decision-making (SDM) tasks, methods like reinforcement learning (RL) and heuristic search have made notable advances in specific cases. However, they often require extensive exploration face challenges generalizing across diverse environments due to their limited grasp of the underlying decision dynamics. contrast, large language models (LLMs) recently emerged as powerful general-purpose tools, capacity maintain vast amounts domain-specific knowledge. To harness this rich...

We address the problem of reward hacking, where maximising a proxy does not necessarily increase true reward. This is key concern for Large Language Models (LLMs), as they are often fine-tuned on human preferences that may accurately reflect objective. Existing work uses various tricks such regularisation, tweaks to model, and hacking detectors, limit influence have model. Luckily, in many contexts medicine, education, law, sparse amount expert data available. In these cases, it unclear...

Learning representations purely from observations concerns the problem of learning a low-dimensional, compact representation which is beneficial to prediction models. Under hypothesis that intrinsic latent factors follow some casual generative models, we argue by causal representation, minimal sufficient causes whole system, can improve robustness and generalization performance machine In this paper, develop method learn such observational data regularizing procedure with mutual information...

Smart growth is an urban planning theory that originated in 1990’s, which has been gradually focused on by researchers. The coincidence with China’s strategies promote urbanization and the transformation of development. goal this paper to apply analytic hierarchy process build evaluation system cities based smart growth. After analyzing factors need be taken into account, we a hierarchical model put different weight compose cities. Finally, use mode for Jinchang city point out scientific...

The capability of imagining internally with a mental model the world is vitally important for human cognition. If machine intelligent agent can learn to create "dream" environment, it then ask what-if questions -- simulate alternative futures that haven't been experienced in past yet and make optimal decisions accordingly. Existing models are established typically by learning spatio-temporal regularities embedded from sensory signal without taking into account confounding factors influence...

This paper studies the location of Wuhan steel logistics distribution center. First all, according to Iron and Steel Plant sales in Hunan Province relative position city, transport costs are calculated from demand point. We further analyze necessity establishing center, using precise center gravity determine actual After establishment total freight is reduced by 15.46 million yuan each city province via year. The results this can provide theoretical basis for node planning related enterprises.

The central idea of smart growth is that structured and strategic planning supports economic growth, addresses community needs, protects the environment.The main assertion this paper to help implementing initiatives into cities' design.Considering three E's sustainability (they are Economically prosperous, socially Equitable, Environmentally Sustainable) ten principles for selects a city as examples obtains data from 2007 2014 comprehensive survey cities managed by National Bureau Statistics...

Top-N recommendation, which aims to learn user ranking-based preference, has long been a fundamental problem in wide range of applications. Traditional models usually motivate themselves by designing complex or tailored architectures based on different assumptions. However, the training data recommender system can be extremely sparse and imbalanced, poses great challenges for boosting recommendation performance. To alleviate this problem, paper, we propose reformulate task within causal...