Quality Diversity (QD) has emerged as a powerful alternative optimization paradigm that aims at generating large and diverse collections of solutions, notably with its flagship algorithm MAP-ELITES (ME) which evolves solutions through mutations and crossovers. While very effective for some unstructured problems, early ME implementations relied exclusively on random search to evolve the population of solutions, rendering them notoriously sample-inefficient for high-dimensional problems, such as when evolving neural networks. Follow-up works considered exploiting gradient information to guide the search in order to address these shortcomings through techniques borrowed from either Black-Box Optimization (BBO) or Reinforcement Learning (RL). While mixing RL techniques with ME unlocked state-of-the-art performance for robotics control problems that require a good amount of exploration, it also plagued these ME variants with limitations common among RL algorithms that ME was free of, such as hyperparameter sensitivity, high stochasticity as well as training instability, including when the population size increases as some components are shared across the population in recent approaches. Furthermore, existing approaches mixing ME with RL tend to be tied to a specific RL algorithm, which effectively prevents their use on problems where the corresponding RL algorithm fails. To address these shortcomings, we introduce a flexible framework that allows the use of any RL algorithm and alleviates the aforementioned limitations by evolving populations of agents (whose definition include hyperparameters and all learnable parameters) instead of just policies. We demonstrate the benefits brought about by our framework through extensive numerical experiments on a number of robotics control problems, some of which with deceptive rewards, taken from the QD-RL literature.
相關內容
Disorders of coronary arteries lead to severe health problems such as atherosclerosis, angina, heart attack and even death. Considering the clinical significance of coronary arteries, an efficient computational model is a vital step towards tissue engineering, enhancing the research of coronary diseases and developing medical treatment and interventional tools. In this work, we applied inverse uncertainty quantification to a microscale agent-based arterial tissue model, a component of a multiscale in-stent restenosis model. Inverse uncertainty quantification was performed to calibrate the arterial tissue model to achieve the mechanical response in line with tissue experimental data. Bayesian calibration with bias term correction was applied to reduce the uncertainty of unknown polynomial coefficients of the attractive force function and achieved agreement with the mechanical behaviour of arterial tissue based on the uniaxial strain tests. Due to the high computational costs of the model, a surrogate model based on Gaussian process was developed to ensure the feasibility of the computation.
Traditionally, approximate dynamic programming is employed in dialogue generation with greedy policy improvement through action sampling, as the natural language action space is vast. However, this practice is inefficient for reinforcement learning (RL) due to the sparsity of eligible responses with high action values, which leads to weak improvement sustained by random sampling. This paper presents theoretical analysis and experiments that reveal the performance of the dialogue policy is positively correlated with the sampling size. To overcome this limitation, we introduce a novel dual-granularity Q-function that explores the most promising response category to intervene in the sampling process. Our approach extracts actions based on a grained hierarchy, thereby achieving the optimum with fewer policy iterations. Additionally, we use offline RL and learn from multiple reward functions designed to capture emotional nuances in human interactions. Empirical studies demonstrate that our algorithm outperforms baselines across automatic metrics and human evaluations. Further testing reveals that our algorithm exhibits both explainability and controllability and generates responses with higher expected rewards.
Taxi-demand prediction is an important application of machine learning that enables taxi-providing facilities to optimize their operations and city planners to improve transportation infrastructure and services. However, the use of sensitive data in these systems raises concerns about privacy and security. In this paper, we propose the use of federated learning for taxi-demand prediction that allows multiple parties to train a machine learning model on their own data while keeping the data private and secure. This can enable organizations to build models on data they otherwise would not be able to access. Despite its potential benefits, federated learning for taxi-demand prediction poses several technical challenges, such as class imbalance, data scarcity among some parties, and the need to ensure model generalization to accommodate diverse facilities and geographic regions. To effectively address these challenges, we propose a system that utilizes region-independent encoding for geographic lat-long coordinates. By doing so, the proposed model is not limited to a specific region, enabling it to perform optimally in any area. Furthermore, we employ cost-sensitive learning and various regularization techniques to mitigate issues related to data scarcity and overfitting, respectively. Evaluation with real-world data collected from 16 taxi service providers in Japan over a period of six months showed the proposed system predicted demand level accurately within 1\% error compared to a single model trained with integrated data. The system also effectively defended against membership inference attacks on passenger data.
We study off-policy evaluation (OPE) of contextual bandit policies for large discrete action spaces where conventional importance-weighting approaches suffer from excessive variance. To circumvent this variance issue, we propose a new estimator, called OffCEM, that is based on the conjunct effect model (CEM), a novel decomposition of the causal effect into a cluster effect and a residual effect. OffCEM applies importance weighting only to action clusters and addresses the residual causal effect through model-based reward estimation. We show that the proposed estimator is unbiased under a new condition, called local correctness, which only requires that the residual-effect model preserves the relative expected reward differences of the actions within each cluster. To best leverage the CEM and local correctness, we also propose a new two-step procedure for performing model-based estimation that minimizes bias in the first step and variance in the second step. We find that the resulting OffCEM estimator substantially improves bias and variance compared to a range of conventional estimators. Experiments demonstrate that OffCEM provides substantial improvements in OPE especially in the presence of many actions.
The generalized linear system (GLS) has been widely used in wireless communications to evaluate the effect of nonlinear preprocessing on receiver performance. Generalized approximation message passing (AMP) is a state-of-the-art algorithm for the signal recovery of GLS, but it was limited to measurement matrices with independent and identically distributed (IID) elements. To relax this restriction, generalized orthogonal/vector AMP (GOAMP/GVAMP) for unitarily-invariant measurement matrices was established, which has been proven to be replica Bayes optimal in uncoded GLS. However, the information-theoretic limit of GOAMP/GVAMP is still an open challenge for arbitrary input distributions due to its complex state evolution (SE). To address this issue, in this paper, we provide the achievable rate analysis of GOAMP/GVAMP in GLS, establishing its information-theoretic limit (i.e., maximum achievable rate). Specifically, we transform the fully-unfolded state evolution (SE) of GOAMP/GVAMP into an equivalent single-input single-output variational SE (VSE). Using the VSE and the mutual information and minimum mean-square error (I-MMSE) lemma, the achievable rate of GOAMP/GVAMP is derived. Moreover, the optimal coding principle for maximizing the achievable rate is proposed, based on which a kind of low-density parity-check (LDPC) code is designed. Numerical results verify the achievable rate advantages of GOAMP/GVAMP over the conventional maximum ratio combining (MRC) receiver based on the linearized model and the BER performance gains of the optimized LDPC codes (0.8~2.8 dB) compared to the existing methods.
Effective exploration is believed to positively influence the long-term user experience on recommendation platforms. Determining its exact benefits, however, has been challenging. Regular A/B tests on exploration often measure neutral or even negative engagement metrics while failing to capture its long-term benefits. To address this, we present a systematic study to formally quantify the value of exploration by examining its effects on the content corpus, a key entity in the recommender system that directly affects user experiences. Specifically, we introduce new metrics and the associated experiment design to measure the benefit of exploration on the corpus change, and further connect the corpus change to the long-term user experience. Furthermore, we investigate the possibility of introducing the Neural Linear Bandit algorithm to build an exploration-based ranking system, and use it as the backbone algorithm for our case study. We conduct extensive live experiments on a large-scale commercial recommendation platform that serves billions of users to validate the new experiment designs, quantify the long-term values of exploration, and to verify the effectiveness of the adopted neural linear bandit algorithm for exploration.
A Low-rank Spectral Optimization Problem (LSOP) minimizes a linear objective subject to multiple two-sided linear matrix inequalities intersected with a low-rank and spectral constrained domain set. Although solving LSOP is, in general, NP-hard, its partial convexification (i.e., replacing the domain set by its convex hull) termed "LSOP-R," is often tractable and yields a high-quality solution. This motivates us to study the strength of LSOP-R. Specifically, we derive rank bounds for any extreme point of the feasible set of LSOP-R and prove their tightness for the domain sets with different matrix spaces. The proposed rank bounds recover two well-known results in the literature from a fresh angle and also allow us to derive sufficient conditions under which the relaxation LSOP-R is equivalent to the original LSOP. To effectively solve LSOP-R, we develop a column generation algorithm with a vector-based convex pricing oracle, coupled with a rank-reduction algorithm, which ensures the output solution satisfies the theoretical rank bound. Finally, we numerically verify the strength of the LSOP-R and the efficacy of the proposed algorithms.
Although numerous clustering algorithms have been developed, many existing methods still leverage k-means technique to detect clusters of data points. However, the performance of k-means heavily depends on the estimation of centers of clusters, which is very difficult to achieve an optimal solution. Another major drawback is that it is sensitive to noise and outlier data. In this paper, from manifold learning perspective, we rethink k-means and present a new clustering algorithm which directly detects clusters of data without mean estimation. Specifically, we construct distance matrix between data points by Butterworth filter such that distance between any two data points in the same clusters equals to a small constant, while increasing the distance between other data pairs from different clusters. To well exploit the complementary information embedded in different views, we leverage the tensor Schatten p-norm regularization on the 3rd-order tensor which consists of indicator matrices of different views. Finally, an efficient alternating algorithm is derived to optimize our model. The constructed sequence was proved to converge to the stationary KKT point. Extensive experimental results indicate the superiority of our proposed method.
Complex event processing (CEP) is a powerful and increasingly more important tool to analyse data streams for Internet of Things (IoT) applications. These data streams often contain private information that requires proper protection. However, privacy protection in CEP systems is still in its infancy, and most existing privacy-preserving mechanisms (PPMs) are adopted from those designed for data streams. Such approaches undermine the quality of the entire data stream and limit the performance of IoT applications. In this paper, we attempt to break the limitation and establish a new foundation for PPMs of CEP by proposing a novel pattern-level differential privacy (DP) guarantee. We introduce two PPMs that guarantee pattern-level DP. They operate only on data that correlate with private patterns rather than on the entire data stream, leading to higher data quality. One of the PPMs provides adaptive privacy protection and brings more granularity and generalization. We evaluate the performance of the proposed PPMs with two experiments on a real-world dataset and on a synthetic dataset. The results of the experiments indicate that our proposed privacy guarantee and its PPMs can deliver better data quality under equally strong privacy guarantees, compared to multiple well-known PPMs designed for data streams.
Classic algorithms and machine learning systems like neural networks are both abundant in everyday life. While classic computer science algorithms are suitable for precise execution of exactly defined tasks such as finding the shortest path in a large graph, neural networks allow learning from data to predict the most likely answer in more complex tasks such as image classification, which cannot be reduced to an exact algorithm. To get the best of both worlds, this thesis explores combining both concepts leading to more robust, better performing, more interpretable, more computationally efficient, and more data efficient architectures. The thesis formalizes the idea of algorithmic supervision, which allows a neural network to learn from or in conjunction with an algorithm. When integrating an algorithm into a neural architecture, it is important that the algorithm is differentiable such that the architecture can be trained end-to-end and gradients can be propagated back through the algorithm in a meaningful way. To make algorithms differentiable, this thesis proposes a general method for continuously relaxing algorithms by perturbing variables and approximating the expectation value in closed form, i.e., without sampling. In addition, this thesis proposes differentiable algorithms, such as differentiable sorting networks, differentiable renderers, and differentiable logic gate networks. Finally, this thesis presents alternative training strategies for learning with algorithms.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191