High-dimensional Continuous Spaces Research Articles

Brief category learning distorts perceptual space for complex scenes.

The formation of categories is known to distort perceptual space: representations are pushed away from category boundaries and pulled toward categorical prototypes. This phenomenon has been studied with artificially constructed objects, whose feature dimensions are easily defined and manipulated. How such category-induced perceptual distortions arise for complex, real-world scenes, however, remains largely unknown due to the technical challenge of measuring and controlling scene features. We address this question by generating realistic scene images from a high-dimensional continuous space using generative adversarial networks and using the images as stimuli in a novel learning task. Participants learned to categorize the scene images along arbitrary category boundaries and later reconstructed the same scenes from memory. Systematic biases in reconstruction errors closely tracked each participant's subjective category boundaries. These findings suggest that the perception of global scene properties is warped to align with a newly learned category structure after only a brief learning experience.

Unsupervised soft-to-hard hashing with contrastive learning

In order to optimize a binary hash code with deep neural networks, greedy optimization with the discrete constraint relaxation is widely used due to its efficiency. However, noises from skipping gradients may destabilize the optimization in contrastive learning since the noises are superimposed on embedding from two distorted views. To reduce such undesirable impact from the noises, we propose a novel soft-to-hard hashing method by introducing an auxiliary loss in the penultimate block of the model based on the information bottleneck principle, which not only propagates the correlations between positives but also makes the embedding to be non-redundant to the samples in a high-dimensional continuous space. Our method beats the state-of-the-art result on three public datasets and demonstrates how soft code can help greedy back-propagation find a better solution during optimization. Another benefit of our method is that it enables a joint training of unsupervised representational learning and hash code generation, and obtains 5.4% mAP gain over conventional step-by-step training on the CIFAR-10 dataset.

Parameters tuning of multi-model database based on deep reinforcement learning

As we all know, the performance of database management system is directly linked to a vast array of knobs, which control various aspects of system operation, ranging from memory and thread counts settings to I/O optimization. Improper settings of configuration parameters are shown to have detrimental effects on performance, reliability and availability of the overall database management system. This is also true for multi-model databases, which use a single platform to support multiple data models. Existing approaches for automatic DBMS knobs tuning are not directly applicable to multi-model databases due to the diversity of multi-model database instances and workloads. Firstly, in cloud environment, they have difficulty adapting to changing environments and diverse workloads. Secondly, they rely on large-scale high-quality training samples that are difficult to obtain. Finally, they focus primarily on throughput metrics, ignoring tuning requirements for resource utilization. Therefore, in this paper, we propose a multi-model database configuration parameters tuning solution named MMDTune. It selects influential parameters, recommends the optimal configurations in a high-dimensional continuous space. For different workloads, the TD3 algorithm is improved to generate reasonable parameter adjustment plans according to the internal state of the multi-model databases. We conduct extensive experiments under 5 different workloads on real cloud databases to evaluate MMDTune. Experimental results show that MMDTune adapts well to a new hardware environment or workloads, and significantly outperforms the representative tuning tools, such as OtterTune, CDBTune.

Novel Random Key Encoding Schemes for the Differential Evolution of Permutation Problems

Differential evolution is a powerful nature-inspired real-parameter optimization algorithm that has been successfully used to solve a number of hard optimization problems. It has been used to tackle both continuous and discrete optimization problems. The application of a continuous method to discrete problems involves several challenges, including solution representation and search space–solution space mapping. In this work, we study random key encoding, a popular encoding scheme that is used to represent permutations in high-dimensional continuous spaces. We analyze the search space it constitutes, study its structure and properties, and introduce two novel modifications of the encoding. We investigate the proposed encoding strategies in the context of four variants of the differential evolution algorithm and demonstrate their usefulness for two widespread permutation problems: 1) the linear ordering problem and 2) the traveling salesman problem.

Hbox {CDBTune}^{+}: An efficient deep reinforcement learning-based automatic cloud database tuning system

Configuration tuning is vital to optimize the performance of a database management system (DBMS). It becomes more tedious and urgent for cloud databases (CDB) due to diverse database instances and query workloads, which make the job of a database administrator (DBA) very difficult. Existing solutions for automatic DBMS configuration tuning have several limitations. Firstly, they adopt a pipelined learning model but cannot optimize the overall performance in an end-to-end manner. Secondly, they rely on large-scale high-quality training samples which are hard to obtain. Thirdly, existing approaches cannot recommend reasonable configurations for a large number of knobs to tune whose potential values live in such high-dimensional continuous space. Lastly, in cloud environments, existing approaches can hardly cope with the changes of hardware configurations and workloads, and have poor adaptability. To address these challenges, we design an end-to-end automatic CDB tuning system, {texttt {CDBTune}}^{+}, using deep reinforcement learning (RL). {texttt {CDBTune}}^{+} utilizes the deep deterministic policy gradient method to find the optimal configurations in a high-dimensional continuous space. {texttt {CDBTune}}^{+} adopts a trial-and-error strategy to learn knob settings with a limited number of samples to accomplish the initial training, which alleviates the necessity of collecting a massive amount of high-quality samples. {texttt {CDBTune}}^{+} adopts the reward-feedback mechanism in RL instead of traditional regression, which enables end-to-end learning and accelerates the convergence speed of our model and improves the efficiency of online tuning. Besides, we propose effective techniques to improve the training and tuning efficiency of {texttt {CDBTune}}^{+} for practical usage in a cloud environment. We conducted extensive experiments under 7 different workloads on real cloud databases to evaluate {texttt {CDBTune}}^{+}. Experimental results showed that {texttt {CDBTune}}^{+} adapts well to a new hardware environment or workload, and significantly outperformed the state-of-the-art tuning tools and DBA experts.

Iterative Local Voting for Collective Decision-making in Continuous Spaces

Many societal decision problems lie in high-dimensional continuous spaces not amenable to the voting techniques common for their discrete or single-dimensional counterparts. These problems are typically discretized before running an election or decided upon through negotiation by representatives. We propose a algorithm called Iterative Local Voting for collective decision-making in this setting. In this algorithm, voters are sequentially sampled and asked to modify a candidate solution within some local neighborhood of its current value, as defined by a ball in some chosen norm, with the size of the ball shrinking at a specified rate. We first prove the convergence of this algorithm under appropriate choices of neighborhoods to Pareto optimal solutions with desirable fairness properties in certain natural settings: when the voters' utilities can be expressed in terms of some form of distance from their ideal solution, and when these utilities are additively decomposable across dimensions. In many of these cases, we obtain convergence to the societal welfare maximizing solution.We then describe an experiment in which we test our algorithm for the decision of the U.S. Federal Budget on Mechanical Turk with over 2,000 workers, employing neighborhoods defined by various L-Norm balls. We make several observations that inform future implementations of such a procedure.

Traversability-Based RRT*for Planetary Rover Path Planning in Rough Terrain with LIDAR Point Cloud Data

Sampling-based search algorithms such as Rapidly-Exploring Random Trees (RRT) have been utilized for mobile robot path planning and motion planning in high dimensional continuous spaces. This paper presents a path planning method for a planetary exploration rover in rough terrain. The proposed method exploits the framework of a sampling-based search, the optimal RRT (RRT*) algorithm. The terrain geometry used for planning is composed of point cloud data close to continuous space captured by a light detection and ranging (LIDAR) sensor. During the path planning phase, the proposed RRT*algorithm directly samples a point (node) from the LIDAR point cloud data. The path planner then considers the rough terrain traversability of the rover during the tree expansion process of RRT*. This process improves conventional RRT*in that the generated path is safe and feasible for the rover in rough terrain. In this paper, simulation study on the proposed path planning algorithm in various real terrain data confirms its usefulness.

Software for project-based learning of robot motion planning

Motion planning is a core problem in robotics concerned with finding feasible paths for a given robot. Motion planning algorithms perform a search in the high-dimensional continuous space of robot configurations and exemplify many of the core algorithmic concepts of search algorithms and associated data structures. Motion planning algorithms can be explained in a simplified two-dimensional setting, but this masks many of the subtleties and complexities of the underlying problem. We have developed software for project-based learning of motion planning that enables deep learning. The projects that we have developed allow advanced undergraduate students and graduate students to reflect on the performance of existing textbook algorithms and their own variations on such algorithms. Formative assessment has been conducted at three institutions. The core of the software used for this teaching module is also used within the Robot Operating System, a widely adopted platform by the robotics research community. This allows for transfer of knowledge and skills to robotics research projects involving a large variety robot hardware platforms.

Learned parametrized dynamic movement primitives with shared synergies for controlling robotic and musculoskeletal systems

A salient feature of human motor skill learning is the ability to exploit similarities across related tasks. In biological motor control, it has been hypothesized that muscle synergies, coherent activations of groups of muscles, allow for exploiting shared knowledge. Recent studies have shown that a rich set of complex motor skills can be generated by a combination of a small number of muscle synergies. In robotics, dynamic movement primitives are commonly used for motor skill learning. This machine learning approach implements a stable attractor system that facilitates learning and it can be used in high-dimensional continuous spaces. However, it does not allow for reusing shared knowledge, i.e., for each task an individual set of parameters has to be learned. We propose a novel movement primitive representation that employs parametrized basis functions, which combines the benefits of muscle synergies and dynamic movement primitives. For each task a superposition of synergies modulates a stable attractor system. This approach leads to a compact representation of multiple motor skills and at the same time enables efficient learning in high-dimensional continuous systems. The movement representation supports discrete and rhythmic movements and in particular includes the dynamic movement primitive approach as a special case. We demonstrate the feasibility of the movement representation in three multi-task learning simulated scenarios. First, the characteristics of the proposed representation are illustrated in a point-mass task. Second, in complex humanoid walking experiments, multiple walking patterns with different step heights are learned robustly and efficiently. Finally, in a multi-directional reaching task simulated with a musculoskeletal model of the human arm, we show how the proposed movement primitives can be used to learn appropriate muscle excitation patterns and to generalize effectively to new reaching skills.

The Ariadne's clew algorithm

We present a new approach to path planning, called the "Ariadne's clew algorithm". It is designed to find paths in high-dimensional continuous spaces and applies to robots with many degrees of free...

The Ariadne's Clew Algorithm

We present a new approach to path planning, called the ``Ariadne's clew algorithm''. It is designed to find paths in high-dimensional continuous spaces and applies to robots with many degrees of freedom in static, as well as dynamic environments --- ones where obstacles may move. The Ariadne's clew algorithm comprises two sub-algorithms, called SEARCH and EXPLORE, applied in an interleaved manner. EXPLORE builds a representation of the accessible space while SEARCH looks for the target. Both are posed as optimization problems. We describe a real implementation of the algorithm to plan paths for a six degrees of freedom arm in a dynamic environment where another six degrees of freedom arm is used as a moving obstacle. Experimental results show that a path is found in about one second without any pre-processing.