United States

Online solvers for partially observable Markov decision processes have difficulty scaling to problems with large action spaces. Monte Carlo tree search with progressive widening attempts to improve scaling by sampling from the action space to construct a policy search tree. The performance of progressive widening search is dependent upon the action sampling policy, often requiring problem-specific samplers. In this work, we present a general method for efficient action sampling based on Bayesian optimization. The proposed method uses a Gaussian process to model a belief over the action-value function and selects the action that will maximize the expected improvement in the optimal action value. We implement the proposed approach in a new online tree search algorithm called Bayesian Optimized Monte Carlo Planning (BOMCP). Several experiments show that BOMCP is better able to scale to large action space POMDPs than existing state-of-the-art tree search solvers.

AAAI 2021

Bayesian Optimized Monte Carlo Planning

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

technical paper

We propose a novel convolutional neural network (ConvNet) equipped with two new semantic calibration and refinement approaches for automatic polyp segmentation from colonoscopy videos. While ConvNets set state-of-the-are performance for this task, it is still difficult to achieve satisfactory results in a real-time manner, which is a necessity in clinical practice. The main obstacle is the huge semantic gap between high-level features and low-level features, making it difficult to take full advantage of complementary semantic information contained in these hierarchical features. Compared with existing solutions, which either directly aggregate these features without considering the semantic gap or employ sophisticated non-local modeling techniques to refine semantic information by introduce many extra computational costs, the proposed ConvNet is able to more precisely yet efficiently calibrate and refine semantic information for better segmentation performance without increasing model complexity; we call the proposed ConvNet as SCR-Net, which has two key modules. We first propose a semantic calibration module (SCM) to effectively transmit the semantic information from high-level layers to low-level layers by learning the semantic-spatial relations during the training procedure. We then propose a semantic refinement module (SRM) to, based on the features calibrated by SCM, enhance the discrimination capability of the features for targeting objects. Extensive experiments on the Kvasir-SEG dataset demonstrate that the proposed SCR-Net is capable of achieving better segmentation accuracy than state-of-the-art approaches with a faster speed. The proposed techniques are general enough to be applied to similar applications where precise and efficient multi-level feature fusion is critical. The code is available at https://github.com/jiafuz/SCR-Net.

Precise Yet Efficient Semantic Calibration and Refinement in ConvNets for Real-Time Polyp Segmentation from Colonoscopy Videos

We present a new deep learning framework for low-contrast image enhancement, which trains the network using the multi-exposure sequences rather than explicit ground-truth images. The purpose of our method is to enhance a low-contrast image so as to contain abundant details in various exposure levels. To realize this, we propose to design the loss function using the structure tensor representation, which has been widely used as high-dimensional image contrast. Our loss function penalizes the difference of the structure tensor between the network output and the multi-exposure images in a multi-scale manner. Eventually, the network trained by the loss function produces a high-quality image approximating the overall contrast of the sequence. We provide in-depth analysis on our method and comparison with conventional loss functions. Quantitative and qualitative evaluations demonstrate that the proposed method outperforms the existing state-of-the-art approaches in various benchmarks.

Deep Low-Contrast Image Enhancement Using Structure Tensor Representation

Context, the embedding of previous collected trajectories, is a powerful construct for Meta-Reinforcement Learning (Meta-RL) algorithms. By conditioning on an effective context, Meta-RL policies can easily generalize to new tasks within a few adaptation steps. We argue that improving the quality of context involves answering two questions: 1. How to train a compact and sufficient encoder that can embed the task-specific information contained in prior trajectories? 2. How to collect informative trajectories of which the corresponding context reflects the specification of tasks? To this end, we propose a novel Meta-RL framework called CCM (Contrastive learning augmented Context-based Meta-RL). We first focus on the contrastive nature behind different tasks and leverage it to train a compact and sufficient context encoder. Further, we train a separate exploration policy and theoretically derive a new information-gain-based objective which aims to collect informative trajectories in a few steps. Empirically, we evaluate our approaches on common benchmarks as well as several complex sparse-reward environments. The experimental results show that CCM outperforms state-of-the-art algorithms by addressing previously mentioned problems respectively.

Towards Effective Context for Meta-Reinforcement Learning: An Approach Based on Contrastive Learning

Modeling how human moves in the space is useful for policy-making in transportation, public safety, and public health. The human movements can be viewed as a dynamic process that human transits between states (e.g., locations) over time. In the human world where intelligent agents like humans or vehicles with human drivers play an important role, the states of agents mostly describe human activities, and the state transition is influenced by both the human decisions and physical constraints from the real-world system (e.g., agents need to spend time to move over a certain distance). Therefore, the modeling of state transition should include the modeling of the agent's decision process and the physical system dynamics. 
In this paper, we propose MoveSD to model state transition in human movement from a novel perspective, by learning the decision model and integrating the system dynamics. MoveSD learns the human movement with Generative Adversarial Imitation Learning and integrates the stochastic constraints from system dynamics in the learning process. To the best of our knowledge, we are the first to learn to model the state transition of moving agents with system dynamics. In extensive experiments on real-world datasets, we demonstrate that the proposed method can generate trajectories similar to real-world ones, and outperform the state-of-the-art methods in predicting the next location and generating long-term future trajectories.

How Do We Move: Modeling Human Movement with System Dynamics

Online solvers for partially observable Markov decision processes have difficulty scaling to problems with large action spaces. This paper proposes a method called PA-POMCPOW to sample a subset of the action space that provides varying mixtures of exploitation and exploration for inclusion in a search tree. The proposed method first evaluates the action space according to a score function that is a linear combination of expected reward and expected information gain. The actions with the highest score are then added to the search tree during tree expansion. Experiments show that PA-POMCPOW is able to outperform existing state-of-the-art solvers on problems with large discrete action spaces.

Improved POMDP Tree Search Planning with Prioritized Action Branching

Solving (mixed) integer (linear) programs, (M)I(L)Ps for short, is a fundamental optimisation task with a wide range of applications in artificial intelligence and computer science in general. While hard in general, recent years have brought about vast progress for solving structurally restricted, (non-mixed) ILPs: n-fold, tree-fold, 2-stage stochastic and multi-stage stochastic programs admit efficient algorithms, and all of these special cases are subsumed by the class of ILPs of small treedepth.

In this paper, we extend this line of work to the mixed case, by showing an algorithm solving MILP in time f(a,d)poly(n), where a is the largest coefficient of the constraint matrix, d is its treedepth, and n is the number of variables.

This is enabled by proving bounds on the denominators (fractionality) of the vertices of bounded-treedepth (non-integer) linear programs. We do so by carefully analysing the inverses of invertible sub-matrices of the constraint matrix. This allows us to afford scaling up the mixed program to the integer grid, and applying the known methods for integer programs.

We then trace the limiting boundary of our "bounded fractionality" approach both in terms of going beyond MILP (by allowing non-linear objectives) as well as its usefulness for generalising other important known tractable classes of ILP.  On the positive side, we show that our result can be generalised from MILP to MIP with piece-wise linear separable convex objectives with integer breakpoints. On the negative side, we show that going even slightly beyond such objectives or considering other natural related tractable classes of ILP leads to unbounded fractionality.

Finally, we show that restricting the structure of only the integral variables in the constraint matrix does not yield tractable special cases.

Parameterized Algorithms for MILPS with Small Treedepth

Building extraction from aerial or satellite images has been an important research issue in remote sensing and computer vision domains for decades. Compared with pixel-wise semantic segmentation models that output raster building segmentation map, polygonal building segmentation approaches produce more realistic building polygons that are in the desirable vector format for practical applications. Despite the substantial efforts over recent years, state-of-the-art polygonal building segmentation methods still suffer from several limitations, e.g., (1) relying on a perfect segmentation map to guarantee the vectorization quality; (2) requiring a complex post-processing procedure; (3) generating inaccurate vertices with a fixed quantity, a wrong sequential order, self-intersections, etc. To tackle the above issues, in this paper, we propose a polygonal building segmentation approach and make the following contributions: (1) We design a multi-task segmentation network for joint semantic and geometric learning via three tasks, i.e., pixel-wise building segmentation, multi-class corner prediction, and edge orientation prediction. (2) We propose a simple but effective vertex generation module for transforming the segmentation contour into high-quality polygon vertices. (3) We further propose a polygon refinement network that automatically moves the polygon vertices into more accurate locations. Results on two popular building segmentation datasets demonstrate that our approach achieves significant improvements for both building instance segmentation (with 2% F1-score gain) and polygon vertex prediction (with 6% F1-score gain) compared with current state-of-the-art methods.

Joint Semantic-Geometric Learning for Polygonal Building Segmentation

We study fundamental clustering problems for incomplete data. Specifically, given a set of incomplete d-dimensional vectors (representing rows of a matrix), the goal is to complete the missing vector entries in a way that admits a partitioning of the vectors into at most k clusters with radius or diameter at most r.  We give tight characterizations of the parameterized complexity of these problems with respect to the parameters k, r, and the minimum number of rows and columns needed to cover all the missing entries. We show that the considered problems are fixed-parameter tractable when parameterized by the three parameters combined, and that dropping any of the three parameters results in parameterized intractability.  A byproduct of our results is that, for the complete data setting, all problems under consideration are fixed-parameter tractable parameterized by k+r.

The Parameterized Complexity of Clustering Incomplete Data

In document-level relation extraction (DocRE), graph structure is generally used to encode relation information in the input document to classify the relation category between each entity pair, and has greatly advanced the DocRE task over the past several years. However, the learned graph representation universally models relation information between all entity pairs regardless of whether there are relationships between these entity pairs. Thus, those entity pairs without relationships disperse the attention of the encoder-classifier DocRE for ones with relationships, which may further hind the improvement of DocRE. To alleviate this issue, we propose a novel encoder-classifier-reconstructor model for DocRE. The reconstructor manages to reconstruct the ground-truth path dependencies from the graph representation, to ensure that the proposed DocRE model pays more attention to encode entity pairs with relationships in the training. Furthermore, the reconstructor is regarded as a relationship indicator to assist relation classification in the inference, which can further improve the performance of DocRE model. Experimental results on a large-scale DocRE dataset show that the proposed model can significantly improve the accuracy of relation extraction on a strong heterogeneous graph-based baseline. The code is publicly available at https://github.com/xwjim/DocRE-Rec.

Document-Level Relation Extraction with Reconstruction

Texts appearing in daily scenes that can be recognized by OCR (Optical Character Recognition) tools contain significant information, such as street name, product brand and prices. Two tasks -- text-based visual question answering and text-based image captioning, with a text extension from existing vision-language applications, are catching on rapidly. 
To address these problems, many sophisticated multi-modality encoding frameworks (such as heterogeneous graph structure) are being used. In this paper, we argue that a simple attention mechanism can do the same or even better job without any bells and whistles. Under this mechanism, we simply split OCR token features into separate visual- and linguistic-attention branches, and send them to a popular Transformer decoder to generate answers or captions. Surprisingly, we find this simple baseline model is rather strong -- it consistently outperforms state-of-the-art (SOTA) models on two popular benchmarks, TextVQA and all three tasks of ST-VQA, although these SOTA models use far more complex encoding mechanisms. Transferring it to text-based image captioning, we also surpass the TextCaps Challenge 2020 winner. We wish this work to set the new baseline for these two OCR text related applications and to inspire  new thinking of multi-modality encoder design. Code is available at https://github.com/ZephyrZhuQi/ssbaseline

Downloads

Next from AAAI 2021

Precise Yet Efficient Semantic Calibration and Refinement in ConvNets for Real-Time Polyp Segmentation from Colonoscopy Videos

Similar lecture

Improved POMDP Tree Search Planning with Prioritized Action Branching

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Precise Yet Efficient Semantic Calibration and Refinement in ConvNets for Real-Time Polyp Segmentation from Colonoscopy Videos

Similar lecture

Improved POMDP Tree Search Planning with Prioritized Action Branching

Downloads