Task (computing)

About: Task (computing) is a research topic. Over the lifetime, 9718 publications have been published within this topic receiving 129364 citations.

An Augmentation System for Fine Manipulation

Abstract: Augmented surgical manipulation tasks can be viewed as a sequence of smaller, simpler steps driven primarily by the surgeon’s input. These steps can be abstracted as controlled interaction of the tool/end-effector with the environment. The basic research problem here is performing a sequence of control primitives. In computing terms, each of the primitives is a predefined computational routine (e.g. compliant motion or some other “macro”) with initiation and termination predicates. The sequencing of these primitives depends upon user control and effects of the environmental interaction. We explore a sensor driven system to perform simple manipulation tasks. The system is composed of a core set of “safe” system states and task specific states and transitions. Using the “steady hand” robot as the experimental platform we investigate using such a system.

Offline Progress of Console Game via Portable Device

Abstract: Methods systems and computer programs method for playing a game using a portable device are presented. One method includes operations for establishing a connection between a portable device and a computing system executing a game, and for receiving a task at the portable device from the computing system. Further, the portable device is disconnected from the computing system allowing the task to be performed using the portable device independently from the computing system. The method includes interaction of a user with the portable device to receive at the portable device input from the user to perform the task. Once the task is performed, the portable device reports that the task has been performed to the computing system. As a result, the game being executed in the computing system progresses after the portable device reports that the task has been performed.

Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theory

Abstract: In meta-learning an agent extracts knowledge from observed tasks, aiming to facilitate learning of novel future tasks. Under the assumption that future tasks are 'related' to previous tasks, the accumulated knowledge should be learned in a way which captures the common structure across learned tasks, while allowing the learner sufficient flexibility to adapt to novel aspects of new tasks. We present a framework for meta-learning that is based on generalization error bounds, allowing us to extend various PAC-Bayes bounds to meta-learning. Learning takes place through the construction of a distribution over hypotheses based on the observed tasks, and its utilization for learning a new task. Thus, prior knowledge is incorporated through setting an experience-dependent prior for novel tasks. We develop a gradient-based algorithm which minimizes an objective function derived from the bounds and demonstrate its effectiveness numerically with deep neural networks. In addition to establishing the improved performance available through meta-learning, we demonstrate the intuitive way by which prior information is manifested at different levels of the network.

Fix the code. Don't tweak the hardware: A new compiler approach to Voltage-Frequency scaling

Abstract: Traditional compiler approaches to optimize power efficiency aim to adjust voltage and frequency at runtime to match the code characteristics to the hardware (e.g., running memory-bound phases at a lower frequency). However, such approaches are constrained by three factors: (i) voltage-frequency transitions are too slow to be applied at instruction granularity, (ii) larger code regions are seldom unequivocally memory- or compute-bound, and, (iii) the available voltage scaling range for future technologies is rapidly shrinking. These factors necessitate new approaches to address power-efficiency at the code-generation level. This paper proposes one such approach to automatically generate power-efficient code using a decoupled access/execute (DAE) model.In DAE a program is split into tasks, where each task consists of two sufficiently coarse-grained phases to enable effective Dynamic Voltage Frequency Scaling (DVFS): (i) the access-phase for data prefetch (heavily memory-bound), and (ii) the execute-phase that performs the actual computation (heavily compute-bound). Our contribution is to provide a compiler methodology to automatically generate the access-phases for a task-based programming system. Our approach is capable of handling both affine (through a polyhedral analysis) and non-affine codes (through optimized task skeletons). Our evaluation shows that the automatically generated versions improve EDP by 25% on average compared to a coupled execution, without any performance degradation, and surpasses the EDP savings of the corresponding hand-crafted tasks by 5%.

Deep Reinforcement Learning with a Combinatorial Action Space for Predicting Popular Reddit Threads

Abstract: We introduce an online popularity prediction and tracking task as a benchmark task for reinforcement learning with a combinatorial, natural language action space. A specified number of discussion threads predicted to be popular are recommended, chosen from a fixed window of recent comments to track. Novel deep reinforcement learning architectures are studied for effective modeling of the value function associated with actions comprised of interdependent sub-actions. The proposed model, which represents dependence between sub-actions through a bi-directional LSTM, gives the best performance across different experimental configurations and domains, and it also generalizes well with varying numbers of recommendation requests.