We propose a task-oriented multi-objective scheduling method based on ant colony optimization (MOSACO) to optimize the finite pool of public and private computing resources in a hybrid cloud computing environment according to deadline and cost constraints. MOSACO is employed to minimize task completion times and costs using time-first and cost-first single-objective optimization strategies, respectively, and to maximize user quality of service and the profit of resource providers using an entropy optimization model. The effectiveness of the MOSACO algorithm based on multiple considerations of task completion time, cost, number of deadline violations, and degree of private resource utilization is verified using simulation and three application examples. Comparisons with similar scheduling methods demonstrate that MOSACO provides the highest optimality, and that the time-first and cost-first strategies provide definite advantages for minimizing completion time and cost, respectively.

Few-shot learning has recently generated increasing popularity in machine learning, which addresses the fundamental yet challenging problem of learning to adapt to new tasks with the limited data. In this paper, we propose a new probabilistic framework that learns to fast adapt with external memory. We model the classifier parameters as distributions that are inferred from the support set and directly applied to the query set for prediction. The model is optimized by formulating as a variational inference problem. The probabilistic modeling enables better handling prediction uncertainty due to the limited data. We impose a discriminative constraint on the feature representations by exploring the class structure, which can improve the classification performance. We further introduce a memory unit to store task-specific information extracted from the support set and used for the query set to achieve explicit adaption to individual tasks. By episodic training, the model learns to acquire the capability of adapting to specific tasks, which guarantees its performance on new related tasks. We conduct extensive experiments on widely-used benchmarks for few-shot recognition. Our method achieves new state-of-the-art performance and largely surpassing previous methods by large margins. The ablation study further demonstrates the effectiveness of the proposed discriminative learning and memory unit.

Few-shot image classification aims to recognize unseen classes with few labeled samples. Existing meta-learning models learn good representations or model parameters to adapt to new tasks with a few training samples. However, when there a domain gap exists between training and test tasks, the learned ability often does not generalize well across domains, resulting in degraded performance on new tasks. In this paper, we propose variational neuron shifting to generate adapted feature representations for few-shot learning. We introduce a working memory module to store the shifted neurons from the support set, which will be accessed to generate adapted feature representations of query samples. Under the meta-learning paradigm, the model is learned to acquire the ability of adaptation with a single sample at meta-training time to further adapt itself to each single test sample at meta-test time. We formulate the adaptation process as a variational Bayesian inference problem, which incorporates the test sample as the condition into the generation of the model neuron shifting. We conduct extensive experiments on both within- and across-domain few-shot classification tasks. Good performance substantiates the effectiveness of the proposed variational neuron shifting, and thorough ablation studies demonstrate the benefit of each component in the proposed model.

The focus of few shot learning research has been on the development of meta-learning recently, where a meta-learner is trained on a variety of tasks in hopes of being generalizable to new tasks. Tasks in meta training and meta test are usually assumed to be from the same domain, which would not necessarily hold in real world scenarios. In this paper, we propose variational hyperparameter inference for few-shot learning across domains. Based on an especially successful algorithm named model agnostic meta learning, the proposed variational hyperparameter inference integrates meta learning and variational inference into the optimization of hyperparameters, which enables the meta-learner with adaptivity for generalization across domains. In particular, we choose to learn adaptive hyperparameters including the learning rate and weight decay to avoid the failure in the face of few labeled examples across domain. Moreover, we model hyperparameters as distributions instead of fixed values, which will further enhance the generalization ability by capturing the uncertainty. Extensive experiments are conducted on two benchmark datasets including few shot learning dataset within-domain and across-domain. The results demonstrate that our methods outperforms previous approaches consistently, and comprehensive ablation studies further validate its effectiveness on few shot learning both within domains and across domains.

Traffic flow prediction is fundamental to traffic information services, control, and guidance. The challenge to accurately model the traffic flow is to comprehensively capture dynamic, global spatial, and local temporal correlations. To address challenges in spatiotemporal dependencies, global similarity, and local dynamics, we propose a multilayer spatiotemporal correlation-aware graph attention network (MSTC-GAT) for traffic flow prediction. Our model contains a multilayer spatial structure-aware module [spatial graph attention network (S-GAT)] using a spatial GAT with hierarchical attention masks and a path-based node correlation matrix to effectively capture local and global spatial dependencies. The temporal structure-aware module [temporal graph attention networks (T-GATs)] constructs a short-term similarity matrix of nodes for the temporal GAT to capture local dynamic temporal dependencies. Finally, a spatiotemporal Transformer (ST-Transformer) fuses weighted spatiotemporal node embeddings to capture global dynamic dependencies for accurate prediction. We conduct extensive experiments on four public benchmark datasets compared with 10 state-of-the-art models. The experimental results demonstrate that the MSTC-GAT outperforms all comparisons for short- and long-term predictions.