IJCAI 2020 Papers

We show that adversarial reinforcement learning (ARL) can be used to produce market marking agents that are robust to adversarial and adaptively-chosen market conditions. To apply ARL, we turn the well-studied single-agent model of Avellaneda and Stoikov [2008] into a discrete-time zero-sum game between a market maker and adversary. The adversary acts as a proxy for other market participants that would like to profit at the market maker's expense. We empirically compare two conventional single-agent RL agents with ARL, and show that our ARL approach leads to: 1) the emergence of risk-averse behaviour without constraints or domain-specific penalties; 2) significant improvements in performance across a set of standard metrics, evaluated with or without an adversary in the test environment, and; 3) improved robustness to model uncertainty. We empirically demonstrate that our ARL method consistently converges, and we prove for several special cases that the profiles that we converge to correspond to Nash equilibria in a simplified single-stage game.

For multi-channel speech recognition, speech enhancement techniques such as denoising or dereverberation are conventionally applied as a front-end processor. Deep learning-based front-ends using such techniques require aligned clean and noisy speech pairs which are generally obtained via data simulation. Recently, several joint optimization techniques have been proposed to train the front-end without parallel data within an end-to-end automatic speech recognition (ASR) scheme. However, the ASR objective is sub-optimal and insufficient for fully training the front-end, which still leaves room for improvement. In this paper, we propose a novel approach which incorporates flow-based density estimation for the robust front-end using non-parallel clean and noisy speech. Experimental results on the CHiME-4 dataset show that the proposed method outperforms the conventional techniques where the front-end is trained only with ASR objective.

The use of reinforcement learning in algorithmic trading is of growing interest, since it offers the opportunity of making profit through the development of autonomous artificial traders, that do not depend on hard-coded rules. In such a framework, keeping uncertainty under control is as important as maximizing expected returns. Risk aversion has been addressed in reinforcement learning through measures related to the distribution of returns. However, in trading it is essential to keep under control the risk of portfolio positions in the intermediate steps. In this paper, we define a novel measure of risk, which we call reward volatility, consisting of the variance of the rewards under the state-occupancy measure. This new risk measure is shown to bound the return variance so that reducing the former also constrains the latter. We derive a policy gradient theorem with a new objective function that exploits the mean-volatility relationship. Furthermore, we adapt TRPO, the well-known policy gradient algorithm with monotonic improvement guarantees, in a risk-averse manner. Finally, we test the proposed approach in two financial environments using real market data.

The guaranteed loan is a debt obligation promise that if one corporation gets trapped in risks, its guarantors will back the loan. When more and more companies involve, they subsequently form complex networks. Detecting and predicting risk guarantee in these networked-loans is important for the loan issuer. Therefore, in this paper, we propose a dynamic graph-based attention neural network for risk guarantee relationship prediction (DGANN). In particular, each guarantee is represented as an edge in dynamic loan networks, while companies are denoted as nodes. We present an attention-based graph neural network to encode the edges that preserve the financial status as well as network structures. The experimental result shows that DGANN could significantly improve the risk prediction accuracy in both the precision and recall compared with state-of-the-art baselines. We also conduct empirical studies to uncover the risk guarantee patterns from the learned attentional network features. The result provides an alternative way for loan risk management, which may inspire more work in the future.

Especially in data-intensive settings, a promising reasoning approach for description logics (DLs) is to rewrite DL theories into sets of rules. Although many such approaches have been considered in the literature, there are still various relevant DLs for which no small rewriting (of polynomial size) is known. We therefore develop small rewritings for the DL \ALCHIQ -- featuring disjunction, number restrictions, and inverse roles -- to disjunctive Datalog. By admitting existential quantifiers in rule heads, we can improve this result to yield only rules of bounded size, a property that is common to all rewritings that were implemented in practice so far.

A major challenge in reinforcement learning is exploration, when local dithering methods such as epsilon-greedy sampling are insufficient to solve a given task. Many recent methods have proposed to intrinsically motivate an agent to seek novel states, driving the agent to discover improved reward. However, while state-novelty exploration methods are suitable for tasks where novel observations correlate well with improved reward, they may not explore more efficiently than epsilon-greedy approaches in environments where the two are not well-correlated. In this paper, we distinguish between exploration tasks in which seeking novel states aids in finding new reward, and those where it does not, such as goal-conditioned tasks and escaping local reward maxima. We propose a new exploration objective, maximizing the reward prediction error (RPE) of a value function trained to predict extrinsic reward. We then propose a deep reinforcement learning method, QXplore, which exploits the temporal difference error of a Q-function to solve hard exploration tasks in high-dimensional MDPs. We demonstrate the exploration behavior of QXplore on several OpenAI Gym MuJoCo tasks and Atari games and observe that QXplore is comparable to or better than a baseline state-novelty method in all cases, outperforming the baseline on tasks where state novelty is not well-correlated with improved reward.

We revisit the notion of i-extension, i.e., the adaption of the fundamental notion of extension to the case of incomplete Abstract Argumentation Frameworks. We show that the definition of i-extension raises some concerns in the "possible" variant, e.g., it allows even conflicting arguments to be collectively considered as members of an (i-)extension. Thus, we introduce the alternative notion of i*-extension overcoming the highlighted problems, and provide a thorough complexity characterization of the corresponding verification problem. Interestingly, we show that the revisitation not only has beneficial effects for the semantics, but also for the complexity: under various semantics, the verification problem under the possible perspective moves from NP-complete to P.

A compelling approach to complex question answering is to convert the question to a sequence of actions, which can then be executed on the knowledge base to yield the answer, aka the programmer-interpreter approach. Use similar training questions to the test question, meta-learning enables the programmer to adapt to unseen questions to tackle potential distributional biases quickly. However, this comes at the cost of manually labeling similar questions to learn a retrieval model, which is tedious and expensive. In this paper, we present a novel method that automatically learns a retrieval model alternately with the programmer from weak supervision, i.e., the system’s performance with respect to the produced answers. To the best of our knowledge, this is the first attempt to train the retrieval model with the programmer jointly. Our system leads to state-of-the-art performance on a large-scale task for complex question answering over knowledge bases. We have released our code at https://github.com/DevinJake/MARL.

We propose a general method for generating counterfactual explanations (CFXs) for a range of Bayesian Network Classifiers (BCs), e.g. single- or multi-label, binary or multidimensional. We focus on explanations built from relations of (critical and potential) influence between variables, indicating the reasons for classifications, rather than any probabilistic information. We show by means of a theoretical analysis of CFXs’ properties that they serve the purpose of indicating (potentially) pivotal factors in the classification process, whose absence would give rise to different classifications. We then prove empirically for various BCs that CFXs provide useful information in real world settings, e.g. when race plays a part in parole violation prediction, and show that they have inherent advantages over existing explanation methods in the literature.

Portfolio selection is an important yet challenging task in AI for FinTech. One of the key issues is how to represent the non-stationary price series of assets in a portfolio, which is important for portfolio decisions. The existing methods, however, fall short of capturing: 1) the complicated sequential patterns for asset price series and 2) the price correlations among multiple assets. In this paper, under a deep reinforcement learning paradigm for portfolio selection, we propose a novel Relation-aware Transformer (RAT) to handle these aspects. Specifically, being equipped with our newly developed attention modules, RAT is structurally innovated to capture both sequential patterns and asset correlations for portfolio selection. Based on the extracted sequential features, RAT is able to make profitable portfolio decisions regarding each asset via a newly devised leverage operation. Extensive experiments on real-world crypto-currency and stock datasets verify the state-of-the-art performance of RAT.

Value function estimation is an important task in reinforcement learning, i.e., prediction. The Boltzmann softmax operator is a natural value estimator and can provide several benefits. However, it does not satisfy the non-expansion property, and its direct use may fail to converge even in value iteration. In this paper, we propose to update the value function with dynamic Boltzmann softmax (DBS) operator, which has good convergence property in the setting of planning and learning. Experimental results on GridWorld show that the DBS operator enables better estimation of the value function, which rectifies the convergence issue of the softmax operator. Finally, we propose the DBS-DQN algorithm by applying the DBS operator, which outperforms DQN substantially in 40 out of 49 Atari games.

Malfunctioning neurons in the brain sometimes operate synchronously, reportedly causing many neurological diseases, e.g. Parkinson’s. Suppression and control of this collective synchronous activity are therefore of great importance for neuroscience, and can only rely on limited engineering trials due to the need to experiment with live human brains. We present the first Reinforcement Learning (RL) gym framework that emulates this collective behavior of neurons and allows us to find suppression parameters for the environment of synthetic degenerate models of neurons. We successfully suppress synchrony via RL for three pathological signaling regimes, characterize the framework’s stability to noise, and further remove the unwanted oscillations by engaging multiple PPO agents.

Face portrait editing has achieved great progress in recent years. However, previous methods either 1) operate on pre-defined face attributes, lacking the flexibility of controlling shapes of high-level semantic facial components (e.g., eyes, nose, mouth), or 2) take manually edited mask or sketch as an intermediate representation for observable changes, but such additional input usually requires extra efforts to obtain. To break the limitations (e.g. shape, mask or sketch) of the existing methods, we propose a novel framework termed r FACE (Reference Guided FAce Component Editing) for diverse and controllable face component editing with geometric changes. Specifically, r-FACE takes an image inpainting model as the backbone, utilizing reference images as conditions for controlling the shape of face components. In order to encourage the framework to concentrate on the target face components, an example-guided attention module is designed to fuse attention features and the target face component features extracted from the reference image. Through extensive experimental validation and comparisons, we verify the effectiveness of the proposed framework.

By leveraging the half-precision floating-point format (FP16) well supported by recent GPUs, mixed precision training (MPT) enables us to train larger models under the same or even smaller budget. However, due to the limited representation range of FP16, gradients can often experience severe underflow problems that hinder backpropagation and degrade model accuracy. MPT adopts loss scaling, which scales up the loss value just before backpropagation starts, to mitigate underflow by enlarging the magnitude of gradients. Unfortunately, scaling once is insufficient: gradients from distinct layers can each have different data distributions and require non-uniform scaling. Heuristics and hyperparameter tuning are needed to minimize these side-effects on loss scaling. We propose gradient scaling, a novel method that analytically calculates the appropriate scale for each gradient on-the-fly. It addresses underflow effectively without numerical problems like overflow and the need for tedious hyperparameter tuning. Experiments on a variety of networks and tasks show that gradient scaling can improve accuracy and reduce overall training effort compared with the state-of-the-art MPT.

Unsupervised image captioning with no annotations is an emerging challenge in computer vision, where the existing arts usually adopt GAN (Generative Adversarial Networks) models. In this paper, we propose a novel memory-based network rather than GAN, named Recurrent Relational Memory Network (R2M). Unlike complicated and sensitive adversarial learning that non-ideally performs for long sentence generation, R2M implements a concepts-to-sentence memory translator through two-stage memory mechanisms: fusion and recurrent memories, correlating the relational reasoning between common visual concepts and the generated words for long periods. R2M encodes visual context through unsupervised training on images, while enabling the memory to learn from irrelevant textual corpus via supervised fashion. Our solution enjoys less learnable parameters and higher computational efficiency than GAN-based methods, which heavily bear parameter sensitivity. We experimentally validate the superiority of R2M than state-of-the-arts on all benchmark datasets.

The Dirichlet Belief Network~(DirBN) has been recently proposed as a promising approach in learning interpretable deep latent representations for objects. In this work, we leverage its interpretable modelling architecture and propose a deep dynamic probabilistic framework -- the Recurrent Dirichlet Belief Network~(Recurrent-DBN) -- to study interpretable hidden structures from dynamic relational data. The proposed Recurrent-DBN has the following merits: (1) it infers interpretable and organised hierarchical latent structures for objects within and across time steps; (2) it enables recurrent long-term temporal dependence modelling, which outperforms the one-order Markov descriptions in most of the dynamic probabilistic frameworks; (3) the computational cost scales to the number of positive links only. In addition, we develop a new inference strategy, which first upward-and-backward propagates latent counts and then downward-and-forward samples variables, to enable efficient Gibbs sampling for the Recurrent-DBN. We apply the Recurrent-DBN to dynamic relational data problems. The extensive experiment results on real-world data validate the advantages of the Recurrent-DBN over the state-of-the-art models in interpretable latent structure discovery and improved link prediction performance.

Partial Multi-label Learning (PML) aims to induce the multi-label predictor from datasets with noisy supervision, where each training instance is associated with several candidate labels but only partially valid. To address the noisy issue, the existing PML methods basically recover the ground-truth labels by leveraging the ground-truth confidence of the candidate label, i.e., the likelihood of a candidate label being a ground-truth one. However, they neglect the information from non-candidate labels, which potentially contributes to the ground-truth label recovery. In this paper, we propose to recover the ground-truth labels, i.e., estimating the ground-truth confidences, from the label enrichment, composed of the relevance degrees of candidate labels and irrelevance degrees of non-candidate labels. Upon this observation, we further develop a novel two-stage PML method, namely Partial Multi-Label Learning with Label Enrichment-Recovery (PML3ER), where in the first stage, it estimates the label enrichment with unconstrained label propagation, then jointly learns the ground-truth confidence and multi-label predictor given the label enrichment. Experimental results validate that PML3ER outperforms the state-of-the-art PML methods.

The goal of Boolean Matrix Factorization (BMF) is to approximate a given binary matrix as the product of two low-rank binary factor matrices, where the product of the factor matrices is computed under the Boolean algebra. While the problem is computationally hard, it is also attractive because the binary nature of the factor matrices makes them highly interpretable. In the last decade, BMF has received a considerable amount of attention in the data mining and formal concept analysis communities and, more recently, the machine learning and the theory communities also started studying BMF. In this survey, we give a concise summary of the efforts of all of these communities and raise some open questions which in our opinion require further investigation.

Electric Vehicle (EV) sharing systems have recently experienced unprecedented growth across the world. One of the key challenges in their operation is vehicle rebalancing, i.e., repositioning the EVs across stations to better satisfy future user demand. This is particularly challenging in the shared EV context, because i) the range of EVs is limited while charging time is substantial, which constrains the rebalancing options; and ii) as a new mobility trend, most of the current EV sharing systems are still continuously expanding their station networks, i.e., the targets for rebalancing can change over time. To tackle these challenges, in this paper we model the rebalancing task as a Multi-Agent Reinforcement Learning (MARL) problem, which directly takes the range and charging properties of the EVs into account. We propose a novel approach of policy optimization with action cascading, which isolates the non-stationarity locally, and use two connected networks to solve the formulated MARL. We evaluate the proposed approach using a simulator calibrated with 1-year operation data from a real EV sharing system. Results show that our approach significantly outperforms the state-of-the-art, offering up to 14% gain in order satisfied rate and 12% increase in net revenue.