In multi-agent systems, complex interacting behaviors arise due to the high correlations among agents. However, previous work on modeling multi-agent interactions from demonstrations is primarily constrained by assuming the independence among policies and their reward structures. In this paper, we cast the multi-agent interactions modeling problem into a multi-agent imitation learning framework with explicit modeling of correlated policies by approximating opponents' policies, which can recover agents' policies that can regenerate similar interactions. Consequently, we develop a Decentralized Adversarial Imitation Learning algorithm with Correlated policies (CoDAIL), which allows for decentralized training and execution. Various experiments demonstrate that CoDAIL can better regenerate complex interactions close to the demonstrators and outperforms state-of-the-art multi-agent imitation learning methods.
Domain adaptation aims to leverage the supervision signal of source domain to obtain an accurate model for target domain, where the labels are not available. To leverage and adapt the label information from source domain, most existing methods employ a feature extracting function and match the marginal distributions of source and target domains in a shared feature space. In this paper, from the perspective of information theory, we show that representation matching is actually an insufficient constraint on the feature space for obtaining a model with good generalization performance in target domain. We then propose variational bottleneck domain adaptation (VBDA), a new domain adaptation method which improves feature transferability by explicitly enforcing the feature extractor to ignore the task-irrelevant factors and focus on the information that is essential to the task of interest for both source and target domains. Extensive experimental results demonstrate that VBDA significantly outperforms state-of-the-art methods across three domain adaptation benchmark datasets.
Sequential recommendation task aims to predict user preference over items in the future given user historical behaviors. The order of user behaviors implies that there are resourceful sequential patterns embedded in the behavior history which reveal the underlying dynamics of user interests. Various sequential recommendation methods are proposed to model the dynamic user behaviors. However, most of the models only consider the user's own behaviors and dynamics, while ignoring the collaborative relations among users and items, i.e., similar tastes of users or analogous properties of items. Without modeling collaborative relations, those methods suffer from the lack of recommendation diversity and thus may have worse performance. Worse still, most existing methods only consider the user-side sequence and ignore the temporal dynamics on the item side. To tackle the problems of the current sequential recommendation models, we propose Sequential Collaborative Recommender (SCoRe) which effectively mines high-order collaborative information using cross-neighbor relation modeling and, additionally utilizes both user-side and item-side historical sequences to better capture user and item dynamics. Experiments on three real-world yet large-scale datasets demonstrate the superiority of the proposed model over strong baselines.
Improving the efficiency of dispatching orders to vehicles is a research hotspot in online ride-hailing systems. Most of the existing solutions for order-dispatching are centralized controlling, which require to consider all possible matches between available orders and vehicles. For large-scale ride-sharing platforms, there are thousands of vehicles and orders to be matched at every second which is of very high computational cost. In this paper, we propose a decentralized execution order-dispatching method based on multi-agent reinforcement learning to address the large-scale order-dispatching problem. Different from the previous cooperative multi-agent reinforcement learning algorithms, in our method, all agents work independently with the guidance from an evaluation of the joint policy since there is no need for communication or explicit cooperation between agents. Furthermore, we use KL-divergence optimization at each time step to speed up the learning process and to balance the vehicles (supply) and orders (demand). Experiments on both the explanatory environment and real-world simulator show that the proposed method outperforms the baselines in terms of accumulated driver income (ADI) and Order Response Rate (ORR) in various traffic environments. Besides, with the support of the online platform of Didi Chuxing, we designed a hybrid system to deploy our model.
Most existing models of multi-agent reinforcement learning (MARL) adopt centralized training with decentralized execution framework. We demonstrate that the decentralized execution scheme restricts agents' capacity to find a better joint policy in team competition games, where each team of agents share the common rewards and cooperate to compete against other teams. To resolve this problem, we propose Signal Instructed Coordination (SIC), a novel coordination module that can be integrated with most existing models. SIC casts a common signal sampled from a pre-defined distribution to team members, and adopts an information-theoretic regularization to encourage agents to exploit in learning the instruction of centralized signals. Our experiments show that SIC can consistently improve team performance over well-recognized MARL models on matrix games and predator-prey games.
GPT-2 and BERT demonstrate the effectiveness of using pre-trained language models (LMs) on various natural language processing tasks. However, LM fine-tuning often suffers from catastrophic forgetting when applied to resource-rich tasks. In this work, we introduce a concerted training framework (\method) that is the key to integrate the pre-trained LMs to neural machine translation (NMT). Our proposed Cnmt consists of three techniques: a) asymptotic distillation to ensure that the NMT model can retain the previous pre-trained knowledge; \item a dynamic switching gate to avoid catastrophic forgetting of pre-trained knowledge; and b)a strategy to adjust the learning paces according to a scheduled policy. Our experiments in machine translation show \method gains of up to 3 BLEU score on the WMT14 English-German language pair which even surpasses the previous state-of-the-art pre-training aided NMT by 1.4 BLEU score. While for the large WMT14 English-French task with 40 millions of sentence-pairs, our base model still significantly improves upon the state-of-the-art Transformer big model by more than 1 BLEU score.
Text-based question answering (TBQA) has been studied extensively in recent years. Most existing approaches focus on finding the answer to a question within a single paragraph. However, many difficult questions require multiple supporting evidence from scattered text among two or more documents. In this paper, we propose Dynamically Fused Graph Network(DFGN), a novel method to answer those questions requiring multiple scattered evidence and reasoning over them. Inspired by human's step-by-step reasoning behavior, DFGN includes a dynamic fusion layer that starts from the entities mentioned in the given query, explores along the entity graph dynamically built from the text, and gradually finds relevant supporting entities from the given documents. We evaluate DFGN on HotpotQA, a public TBQA dataset requiring multi-hop reasoning. DFGN achieves competitive results on the public board. Furthermore, our analysis shows DFGN produces interpretable reasoning chains.
Knowledge base is one of the main forms to represent information in a structured way. A knowledge base typically consists of Resource Description Frameworks (RDF) triples which describe the entities and their relations. Generating natural language description of the knowledge base is an important task in NLP, which has been formulated as a conditional language generation task and tackled using the sequence-to-sequence framework. Current works mostly train the language models by maximum likelihood estimation, which tends to generate lousy sentences. In this paper, we argue that such a problem of maximum likelihood estimation is intrinsic, which is generally irrevocable via changing network structures. Accordingly, we propose a novel Triple-to-Text (T2T) framework, which approximately optimizes the inverse Kullback-Leibler (KL) divergence between the distributions of the real and generated sentences. Due to the nature that inverse KL imposes large penalty on fake-looking samples, the proposed method can significantly reduce the probability of generating low-quality sentences. Our experiments on three real-world datasets demonstrate that T2T can generate higher-quality sentences and outperform baseline models in several evaluation metrics.