Continuous self-adaptive optimization to learn multi-task multi-agent

Multi-agent reinforcement learning (MARL) methods have shown superior performance to solve a variety of real-world problems focusing on learning distinct policies for individual tasks. These approaches face problems when applied to the non-stationary real-world: agents trained in specialized tasks cannot achieve satisfied generalization performance across multiple tasks; agents have to learn and store specialized policies for individual task and reliable identities of tasks are hardly observable in practice. To address the challenge continuously adapting to multiple tasks in MARL, we formalize the problem into a two-stage curriculum. Single-task policies are learned with MARL approaches, after that we develop a gradient-based Self-Adaptive Meta-Learning algorithm, SAML, that cannot only distill single-task policies into a unified policy but also can facilitate the unified policy to continuously adapt to new incoming tasks. In addition, to validate the continuous adaptation performance on complex task, we extend the widely adopted StarCraft benchmark SMAC and develop a new multi-task multi-agent StarCraft environment, Meta-SMAC, for testing various aspects of continuous adaptation method. Our experiments with a population of agents show that our method enables significantly more efficient adaptation than reactive baselines across different scenarios.

Simple contextual cueing prevents retroactive interference in short-term perceptual training of orientation detection tasks

Perceptual training of multiple tasks suffers from interference between the trained tasks. Here, we conducted four psychophysical experiments with separate groups of participants to investigate the possibility of preventing the interference in short-term perceptual training. We trained the participants to detect two orientations of Gabor stimuli in two adjacent days at the same retinal location and examined the interference of training effects between the two orientations. The results showed significant retroactive interference from the second orientation to the first orientation (Experiments 1 and 2). Introducing a 6-hour interval between the pre-test and training of the second orientation did not eliminate the interference effect, excluding the interpretation of disrupted reconsolidation as the pre-test of the second orientation may reactivate and destabilize the representation of the first orientation (Experiment 3). Finally, the training of the two orientations was accompanied by fixations in two colors, each served as a contextual cue for one orientation. The results showed that the retroactive interference was not evident after introducing these passively perceived contextual cues (Experiment 4). Our findings suggest that the retroactive interference effect in short-term perceptual training of orientation detection tasks was likely the result of higher-level factors such as shared contextual cues embedded in the tasks. The effect of multiple perceptual training could be facilitated by associating the trained tasks with different contextual cues.

Rapid Classification and Analysis for E-Commerce Goods Based on Multitask Learning

Today’s E-commerce is hot, while the categorization of goods cannot be handled better, especially to achieve the demand of multiple tasks. In this paper, we propose a multitask learning model based on a CNN in parallel with a BiLSTM optimized by an attention mechanism as a training network for E-commerce. The results showed that the fast classification task of E-commerce was performed using only 10% of the total number of products. The experimental results show that the accuracy of w-item2vec for product classification can be close to 50% with only 10% of the training data. Both models significantly outperform other models in terms of classification accuracy.

Few-Shot Tuning Framework for Automated Terms of Service Generation

In this paper, we introduce BART2S a novel framework based on BART pretrained models to generate terms of service in high quality. The framework contains two parts: a generator finetuned with multiple tasks and a discriminator fine-tuned to distinguish the fair and unfair terms. Besides the novelty in design and the implementation contributions, the proposed framework can support drafting terms of service, a growing need in the digital age. Our proposed approach allows the system to reach a balance between automation and the will expression of the service provider. Through experiments, we demonstrate the effectiveness of the method and discuss potential future directions.

Energy-Efficient Allocation for Multiple Tasks in Mobile Edge Computing

Mobile edge computing (MEC) allows a mobile device to offload tasks to the nearby server for remote execution to enhance the performance of user equipment. A major challenge of MEC is to design an efficient algorithm for task allocation. In contrast to previous work on MEC, which mainly focuses on single-task allocation for a mobile device with only one task to be completed, this paper considers a mobile device with multiple tasks or an application with multiple tasks. This assumption does not hold in real settings because a mobile device may have multiple tasks waiting to execute. We address the problem of task allocation with minimum total energy consumption considering multi-task settings in MEC, in which a mobile device has one or more tasks. We consider the binary computation offloading mode and formulate multi-task allocation as an integer programming problem that is strongly $NP$-hard. We propose an approximation algorithm and show it is a polynomial-time approximation scheme that saves the maximum energy. Therefore, our proposed algorithm achieves a tradeoff between optimality loss and time complexity. We analyze the performance of the proposed algorithm by performing extensive experiments. The results of the experiments demonstrate that our proposed approximation algorithm is capable of finding near-optimal solutions, and achieves a good balance of speed and quality.

Women's Home Office

The study investigated the implications for women from the coast of Paraná, Brazilin regard to the home office in life, family, and work during the COVID-19 pandemic. The research was conducted on 20 women, and 3 (three) categories of analysis were taken into consideration: woman, family, and work. For data analysis, the content of Bardin analysis technique was used. The study revealed that the imposition of the home office brought new challenges for women to reconcile home care, family, and work, which caused an overload of responsibilities leading to fatigue due to the multiple tasks performed; as well as the lack of preparation of companies to adopt the modality, which results in an environment that hinders the performance of their functions and affects the quality of life of workers.

Sensory Organ Investment Varies with Body Size and Sex in the Butterfly Pieris napi

In solitary insect pollinators such as butterflies, sensory systems must be adapted for multiple tasks, including nectar foraging, mate-finding, and locating host-plants. As a result, the energetic investments between sensory organs can vary at the intraspecific level and even among sexes. To date, little is known about how these investments are distributed between sensory systems and how it varies among individuals of different sex. We performed a comprehensive allometric study on males and females of the butterfly Pieris napi where we measured the sizes and other parameters of sensory traits including eyes, antennae, proboscis, and wings. Our findings show that among all the sensory traits measured, only antenna and wing size have an allometric relationship with body size and that the energetic investment in different sensory systems varies between males and females. Moreover, males had absolutely larger antennae and eyes, indicating that they invest more energy in these organs than females of the same body size. Overall, the findings of this study reveal that the size of sensory traits in P. napi are not necessarily related to body size and raises questions about other factors that drive sensory trait investment in this species and in other insect pollinators in general.

Consideration and analysis for attachment position of supernumerary robotic limbs

The supernumerary robotic limbs(SRLs) is a new type of wearable robot that assists the operator with additional robotic limbs and allows the operator to perform multiple tasks simultaneously. Due to the SRLs having various combinations of robotic limb and attachment positions, and there is an insufficient discussion on the influence of different wear positions on the SRLs. Therefore, this paper improved the evaluation indexes from previous studies and presents an experimental evaluation of the performance of indexes between humans and SRLs. This paper analyzed the 5 different positions based on the improved evaluation indexes, 2 optimal positions are found with the simulation experiment. Then the two design factors to improve the performance of evaluation indexes are discussed. The evaluation indexes can be utilized as a design parameter for evaluating human-robot interactions of SRLs.

Brief analysis: development of nuclear submarines and elaboration of industry standards for power plant systems

Object and purpose of research. A wide application of nuclear technologies and new types of weapons, electronic sensors, automation systems and tools have increased combat capabilities of the Navy and its major component represented by submarine forces. Materials and methods. Nuclear submarines are core weapons of the Naval Forces, which can efficiently perform strategic and tactical roles, including multiple tasks. Main results. This paper briefly analyzes the steam-generation and steam-turbine plants development of submarines and elaboration of industry standards used for design of the systems supporting of nuclear submarines. Conclusions. This paper assesses the relevancy of standards used in development of steam-generation and steam-turbine plants for advanced nuclear submarines and identifies the scope for their further improvement to take into account modern nuclear submarine requirements in the design.

Research on resource allocation strategy of data center based on cloud computing

With the development of Internet and information technology, cloud computing has attracted extensive attention from industry and academia. The large scale of resources, concurrent execution of multiple tasks and dynamic changes of application resource requests make the resource allocation of data center face severe challenges. To solve the problem of low balance of traditional resource allocation, this paper focuses on the resource allocation optimization of data center, and proposes the resource allocation strategy of data center based on cloud computing, so as to complete the effective resource allocation and assignment. This paper also verifies the designed resource allocation method through example research. The research shows that the distribution balance degree of resource allocation strategy based on cloud computing is significantly higher than the control group, which proves that the designed resource allocation strategy can solve the problem of low balance of traditional resource allocation.