- A vibrant theoretical research area are efficient exact parameterized algorithms. Very recent solving competitions such as the PACE challenge show that there is also increasing practical interest in the parameterized algorithms community. An important research question is whether dedicated parameterized exact algorithms exhibit certain practical relevance and one can even beat well-established problem solvers. We consider the logic-based declarative modeling language and problem solving framework Answer Set Programming (ASP). State-of-the-art ASP solvers rely considerably on Sat-based algorithms. An ASP solver (DynASP2), which is based on a classical dynamic programming on tree decompositions, has been published very recently. Unfortunately, DynASP2 can outperform modern ASP solvers on programs of small treewidth only if the question of interest is to count the number of solutions. In this paper, we describe underlying concepts of our new implementation (DynASP2.5) that shows competitive behavior to state-of-the-art ASP solvers even for finding just one solution when solving problems as the Steiner tree problem that have been modeled in ASP on graphs with low treewidth. Our implementation is based on a novel approach that we call multi-pass dynamic programming (M-DPSINC).
- Jun 29 2017 cs.AI arXiv:1706.09278v1Learning relations based on evidence from knowledge bases relies on processing the available relation instances. Many relations, however, have clear domain and range, which we hypothesize could help learn a better, more generalizing, model. We include such information in the RESCAL model in the form of a regularization factor added to the loss function that takes into account the types (categories) of the entities that appear as arguments to relations in the knowledge base. We note increased performance compared to the baseline model in terms of mean reciprocal rank and hits@N, N = 1, 3, 10. Furthermore, we discover scenarios that significantly impact the effectiveness of the type regularizer.
- Class-agnostic object tracking is particularly difficult in cluttered environments as target specific discriminative models cannot be learned a priori. Inspired by how the human visual cortex employs spatial attention and separate "where" and "what" processing pathways to actively suppress irrelevant visual features, this work develops a hierarchical attentive recurrent model for single object tracking in videos. The first layer of attention discards the majority of background by selecting a region containing the object of interest, while the subsequent layers tune in on visual features particular to the tracked object. This framework is fully differentiable and can be trained in a purely data driven fashion by gradient methods. To improve training convergence, we augment the loss function with terms for a number of auxiliary tasks relevant for tracking. Evaluation of the proposed model is performed on two datasets of increasing difficulty: pedestrian tracking on the KTH activity recognition dataset and the KITTI object tracking dataset.
- In sequence prediction task, there mainly exist two types of learning algorithms. One genre based on maximum likelihood estimation (MLE) supervises the training in every time step via a surrogate loss function but leads to train-test discrepancy and hard-to-train problem. The other genre based on Reinforcement Learning gives the model freedom to generate samples and guides the training by feeding back task-level reward but suffers from high variance and instability problem. In general, these two frameworks both have their own trade-offs and complement each other. In this paper, we introduce a hybrid-coach framework to combine these two existing algorithms so that the supervision from MLE can stabilize RL training while RL can incorporate task-level loss into MLE training. To augment more easy-to-learn samples for MLE training and denser reward for RL training, the coach is encouraged to be close to both the data and model distribution. During training, the coach acts as a proxy target and pulls the model distribution to the data distribution, as the model makes progress it correspondingly upgrades itself to become harsher and finally anneal to the ultimate target distribution. With the coordination of the coach system, our hybrid-coach framework can efficiently resolve RL's reward sparsity problem and MLE's data sparsity problem. We apply our algorithm on two Machine Translation Task and show significant improvements over the state-of-art systems.
- A descriptive approach for automatic generation of visual blends is presented. The implemented system, the Blender, is composed of two components: the Mapper and the Visual Blender. The approach uses structured visual representations along with sets of visual relations which describe how the elements (in which the visual representation can be decomposed) relate among each other. Our system is a hybrid blender, as the blending process starts at the Mapper (conceptual level) and ends at the Visual Blender (visual representation level). The experimental results show that the Blender is able to create analogies from input mental spaces and produce well-composed blends, which follow the rules imposed by its base-analogy and its relations. The resulting blends are visually interesting and some can be considered as unexpected.