United States

Reward shaping is one of the most effective methods to tackle the crucial yet challenging problem of credit assignment and accelerate Reinforcement Learning. However, designing shaping functions usually requires rich expert knowledge and hand-engineering, and the difficulties are further exacerbated given multiple tasks to solve. In this paper, we consider reward shaping on a distribution of tasks that share state spaces but not necessarily action spaces. We provide insights into optimal reward shaping, and propose a novel meta-learning framework to automatically learn such reward shaping to apply on newly sampled tasks. Theoretical analysis and extensive experiments establish us as the state-of-the-art in learning task-distribution reward shaping, outperforming previous such works (Konidaris and Barto 2006; Snel and Whiteson 2014). We further show that our method outperforms learning intrinsic rewards (Yang et al. 2019; Zheng et al. 2020), outperforms Rainbow (Hessel et al. 2018) in complex pixel-based CoinRun games, and is also better than hand-designed reward shaping on grids. While the goal of this paper is to learn reward shaping rather than to propose new general meta-learning algorithms as PEARL (Rakelly et al. 2019) or MQL (Fakoor et al. 2020), our framework based on MAML (Finn, Abbeel, and Levine 2017) also outperforms PEARL / MQL, and could combine with them for further improvement.

AAAI 2021

Learning Task-Distribution Reward Shaping with Meta-Learning

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

Despite the empirical success in various domains, it has been revealed that deep neural networks are vulnerable to maliciously perturbed input data that much degrade their performance. This is known as adversarial attacks. To counter adversarial attacks, adversarial training formulated as a form of robust optimization has been demonstrated to be effective. However, conducting adversarial training brings much computational overhead compared with standard training. In order to reduce the computational cost, we propose an annealing mechanism, Amata, to reduce the overhead associated with adversarial training. The proposed Amata is provably convergent, well-motivated from the lens of optimal control theory and can be combined with existing acceleration methods to further enhance performance. It is demonstrated that on standard datasets, Amata can achieve similar or better robustness with around 1/3 to 1/2 the computational time compared with traditional methods. In addition, Amata can be incorporated into  other adversarial training acceleration algorithms (e.g. YOPO, Free, Fast, and ATTA), which leads to further reduction in computational time on large-scale problems.

Amata: An Annealing Mechanism for Adversarial Training Acceleration

LiDAR-based 3D object detection is an important task for autonomous driving and current approaches suffer from sparse and partial point clouds caused by distant and occluded objects. In this paper, we propose a novel two-stage framework, namely PC-RGNN, which deals with these challenges by two
specific solutions. On the one hand, we introduce a point cloud completion module to recover high-quality proposals of dense points and entire view with original structures preserved. On the other hand, a graph neural network module, is designed, which comprehensively captures relations among
points by the local-global attention mechanism as well as the multi-scale graph based context aggregation and substantially strengthens encoded features. Extensive experiments on the KITTI benchmark show that the proposed approach outperforms the previous state-of-the-art baselines by remarkable margins, highlighting its effectiveness.

PC-RGNN: Point Cloud Completion and Graph Neural Network for 3D Object Detection

Text generation tasks, including translation, summarization, language models, and etc. see rapid growth during recent years. Despite the remarkable achievements, the repetition problem has been observed in nearly all text generation models undermining the generation performance extensively. To solve the repetition problem, many methods have been proposed, but there is no existing theoretical analysis to show why this problem happens and how it is resolved. In this paper, we propose a new framework for theoretical analysis for the repetition problem. We first define the Average Repetition Probability (ARP) to characterize the repetition problem quantitatively. Then, we conduct an extensive analysis of the Markov generation model and derive several upper bounds of the average repetition probability with intuitive understanding. We show that most of the existing methods are essentially minimizing the upper bounds explicitly or implicitly. Grounded on our theory, we show that the repetition problem is, unfortunately, caused by the traits of our language itself. One major reason is attributed to the fact that there exist too many words predicting the same word as the subsequent word with high probability. Consequently, it is easy to go back to that word and form repetitions and we dub it as the high inflow problem. Furthermore, we extend our analysis to broader generation models by deriving a concentration bound of the average repetition probability for a general generation model. Finally,  based on the theoretical upper bounds, we propose a novel rebalanced encoding approach to alleviate the high inflow problem and thus reducing the upper bound.
    The experimental results show that our theoretical framework is applicable in general generation models and our proposed rebalanced encoding approach alleviates the repetition problem significantly in both the translation task and the language modeling task. The source code of this paper can be obtained from https://github.com/fuzihaofzh/repetition-problem-nlg.

A Theoretical Analysis of the Repetition Problem in Text Generation

Text coherence plays a key role in document quality assessment. Most existing text coherence methods only focus on similarity of adjacent sentences. However, local coherence exists in sentences with broader contexts and diverse rhetoric relations, rather than just adjacent sentences similarity. Besides, the highlevel text coherence is also an important aspect of document quality. To this end, we propose a hierarchical coherence model for document quality assessment. In our model, we implement a local attention mechanism to capture the location semantics, bilinear tensor layer for measure coherence and max-coherence pooling for acquiring high-level coherence. We evaluate the proposed method on two realistic tasks: news quality judgement and automated essay scoring.   Experimental results demonstrate the validity and superiority of our work.

Hierarchical Coherence Modeling for Document Quality Assessment

One significant factor we expect the video representation learning to capture, especially in contrast with the image representation learning, is the object motion. However, we found that in the current mainstream video datasets, some action categories are highly related with the scene where the action happens, making the model tend to degrade to a solution where only the scene information is encoded. For example, a trained model may predict a video as playing football simply because it sees the field, neglecting that the subject is dancing as a cheerleader on the field. This is against our original intention towards the video representation learning and may bring scene bias on a different dataset that can not be ignored. In order to tackle this problem, we propose to decouple the scene and the motion (DSM) with two simple operations, so that the model attention towards the motion information is better paid. Specifically, we construct a positive clip and a negative clip for each video. Compared to the original video, the positive/negative is motion-untouched/broken but scene-broken/untouched by Spatial Local Disturbance and Temporal Local Disturbance. Our objective is to pull the positive closer while pushing the negative farther to the original clip in the latent space. In this way, the impact of the scene is weakened while the temporal sensitivity of the network is further enhanced. We conduct experiments on two tasks with various backbones and different pre-training datasets, and find that our method surpass the SOTA methods with a remarkable 8.1% and 8.8% improvement towards action recognition task on the UCF101 and HMDB51 datasets respectively using the same backbone.

Enhancing Unsupervised Video Representation Learning by Decoupling the Scene and the Motion

Condorcet extensions have long held a prominent place in social choice theory. A Condorcet extension will return the Condorcet winner as the unique winner whenever such an alternative exists. However, the definition of a Condorcet extension does not take into account possible manipulation by the voters. A profile where all agents vote truthfully may have a Condorcet winner, but this alternative may not end up in the set of winners if agents are acting strategically. Focusing on the class of tournament solutions, we show that many natural social choice functions in this class, such as the well-known Copeland and Slater rules, cannot guarantee the preservation of Condorcet winners when agents behave strategically. Our main result in this respect is an impossibility theorem that establishes that no tournament solution satisfying a very weak decisiveness requirement can provide such a guarantee. On the bright side, we identify several indecisive but otherwise attractive tournament solutions that do guarantee the preservation of Condorcet winners under strategic manipulation for a large class of preference extensions.

Preserving Condorcet Winners under Strategic Manipulation

We consider a sequential assortment selection problem where the user choice is given by a multinomial logit (MNL) choice model whose parameters are unknown. In each period, the learning agent observes a d-dimensional contextual information about the user and the N available items, and offers an assortment of size K to the user, and observes the bandit feedback of the item chosen from the assortment. We propose upper confidence bound based algorithms for this MNL contextual bandit. The first algorithm is a simple and practical method that achieves an O(dâˆšT) regret over T rounds. Next, we propose a second algorithm which achieves a O(âˆšdT) regret. This matches the lower bound for the MNL bandit problem, up to logarithmic terms, and improves on the best-known result by a âˆšd factor. To establish this sharper regret bound, we present a non-asymptotic confidence bound for the maximum likelihood estimator of the MNL model that may be of independent interest as its own theoretical contribution. We then revisit the simpler, significantly more practical, first algorithm and show that a simple variant of the algorithm achieves the optimal regret for a broad class of important applications.

Multinomial Logit Contextual Bandits: Provable Optimality and Practicality

Depth adjustment aims to enhance the visual experience of stereoscopic 3D (S3D) images, which accompanied with improving visual comfort and depth perception. For a human expert, the depth adjustment procedure is a sequence of iterative decision making. The human expert iteratively adjusted the depth until he is satisfied with the both levels of visual comfort and the perceived depth. In this work, we present a novel deep reinforcement learning (DRL)-based approach for depth adjustment named VCA-RL (Visual Comfort Aware Reinforcement Learning) to explicitly model human sequential decision making in depth editing operations. We formulate the depth adjustment process as a Markov decision process where actions are defined as camera movement operations to control the distance between the left and right cameras. Our agent is trained based on the guidance of an objective visual comfort assessment metric to learn the optimal sequence of camera movement actions in terms of perceptual aspects in stereoscopic viewing. With extensive experiments and user studies, we show the effectiveness of our VCA-RL model on three different S3D databases.

Visual Comfort Aware-Reinforcement Learning for Depth Adjustment of Stereoscopic 3D Images

Stackelberg equilibrium is a solution concept in two-player games where the leader has commitment rights over the follower. In recent years, it has become a cornerstone of many security applications, including airport patrolling and wildlife poaching prevention. Even though many of these settings are sequential in nature, existing techniques pre-compute the entire solution ahead of time. In this paper, we present a theoretically sound and empirically effective way to apply search, which leverages extra online computation to improve a solution, to the computation of Stackelberg equilibria in general-sum games. Instead of the leader attempting to solve the full game upfront, an approximate "blueprint" solution is first computed offline and is then improved online for the particular subgames encountered in actual play. We prove that our search technique is guaranteed to perform no worse than the pre-computed blueprint strategy, and empirically demonstrate that it enables approximately solving significantly larger games compared to purely offline methods. We also show that our search operation may be cast as a smaller Stackelberg problem, making our method complementary to existing algorithms based on strategy generation.

Safe Search for Stackelberg Equilibria in Extensive-Form Games

Domain generalization (DG) offers a preferable real-world setting for Person Re-Identification (Re-ID), which trains a model using multiple source domain datasets and expects it to perform well in an unseen target domain without any model updating. Unfortunately, most DG approaches are designed explicitly for classification tasks, which fundamentally differs from the retrieval task Re-ID. Moreover, existing applications of DG in Re-ID cannot correctly handle the massive variation among Re-ID datasets. In this paper, we identify two fundamental challenges in DG for Person Re-ID: domain-wise variations and identity-wise similarities. To this end, we propose an end-to-end Dual Distribution Alignment Network (DDAN) to learn domain-invariant features with dual-level constraints: the domain-wise adversarial feature learning and the identity-wise similarity enhancement. These constraints effectively reduce the domain-shift among multiple source domains further while agreeing to real-world scenarios. We evaluate our method in a large-scale DG Re-ID benchmark and compare it with various cutting-edge DG approaches. Quantitative results show that DDAN achieves state-of-the-art performance.

Downloads

Next from AAAI 2021

Amata: An Annealing Mechanism for Adversarial Training Acceleration

Similar lecture

Temporal-Logic-Based Reward Shaping for Continuing Reinforcement Learning Tasks

.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

Amata: An Annealing Mechanism for Adversarial Training Acceleration

Similar lecture

Temporal-Logic-Based Reward Shaping for Continuing Reinforcement Learning Tasks

Downloads