In a multi-agent system, an agent's optimal policy will typically depend on the policies chosen by others. Therefore, a key issue in multi-agent systems research is that of predicting the behaviours of others, and responding promptly to changes in such behaviours. One obvious possibility is for each agent to broadcast their current intention, for example, the currently executed option in a hierarchical reinforcement learning framework. However, this approach results in inflexibility of agents if options have an extended duration and are dynamic. While adjusting the executed option at each step improves flexibility from a single-agent perspective, frequent changes in options can induce inconsistency between an agent's actual behaviour and its broadcast intention. In order to balance flexibility and predictability, we propose a dynamic termination Bellman equation that allows the agents to flexibly terminate their options. We evaluate our model empirically on a set of multi-agent pursuit and taxi tasks, and show that our agents learn to adapt flexibly across scenarios that require different termination behaviours.

While cyclic scheduling is involved in numerous real-world applications, solving the derived problem is still of exponential complexity. This paper focuses specifically on modelling the manufacturing application as a cyclic job shop problem and we have developed an efficient neural network approach to minimise the cycle time of a schedule. Our approach introduces an interesting model for a manufacturing production, and it is also very efficient, adaptive and flexible enough to work with other techniques. Experimental results validated the approach and confirmed our hypotheses about the system model and the efficiency of neural networks for such a class of problems. Keywords Neural networks, Cyclic job scheduling, Manufacturing processes, Unconstrained optimisation problem 690 M.-T.

We present an axiomatic way of assigning probabilities to black box models. In particular, we quantify an upper bound for probability of a model or in terms of information theory, a lower bound for amount of information that is stored in a model. In our setup, maximizing probabilities of models is equivalent to removing assumptions or information stored in the model. Furthermore, we represent the problem of learning from an alternative view where the underlying probability space is considered directly. In this perspective both the true underlying model and the model at hand are events. Consequently, the problem of learning is represented as minimizing the probability of the symmetric difference of the model and the true underlying model.

A salient approach to interpretable machine learning is to restrict modeling to simple and hence understandable models. In the Bayesian framework, this can be pursued by restricting the model structure and prior to favor interpretable models. Fundamentally, however, interpretability is about users' preferences, not the data generation mechanism: it is more natural to formulate interpretability as a utility function. In this work, we propose an interpretability utility, which explicates the trade-off between explanation fidelity and interpretability in the Bayesian framework. The method consists of two steps. First, a reference model, possibly a black-box Bayesian predictive model compromising no accuracy, is constructed and fitted to the training data. Second, a proxy model from an interpretable model family that best mimics the predictive behaviour of the reference model is found by optimizing the interpretability utility function. The approach is model agnostic - neither the interpretable model nor the reference model are restricted to be from a certain class of models - and the optimization problem can be solved using standard tools in the chosen model family. Through experiments on real-word data sets using decision trees as interpretable models and Bayesian additive regression models as reference models, we show that for the same level of interpretability, our approach generates more accurate models than the earlier alternative of restricting the prior. We also propose a systematic way to measure stabilities of interpretabile models constructed by different interpretability approaches and show that our proposed approach generates more stable models.

Redistribution mechanisms have been proposed for more efficient resource allocation but not for profit. We consider redistribution mechanism design for the first time in a setting where participants are connected and the resource owner is only aware of her neighbours. In this setting, to make the resource allocation more efficient, the resource owner has to inform the others who are not her neighbours, but her neighbours do not want more participants to compete with them. Hence, the goal is to design a redistribution mechanism such that participants are incentivized to invite more participants and the resource owner does not earn or lose much money from the allocation. We first show that existing redistribution mechanisms cannot be directly applied in the network setting to achieve the goal. Then we propose a novel network-based redistribution mechanism such that all participants in the network are invited, the allocation is more efficient and the resource owner has no deficit. Introduction The problem of resource allocation has recently caught the public imagination, where the resource owner has to decide the allocation of the item among a group of self-interested agents. Since the valuation differs from agents, it is a natural objective for the owner to pursue the efficiency of the allocation, i.e., allocating the item to the agent with the highest valuation. In many scenarios, the owner does not really aim at making profits but hopes the wealth maintained among the agents. For example, the government wants to build a library in a community that values it most; a charity distributes a donation to the recipient who needs it most; a hospital allocates doctors to rural areas where doctors are highly demanded. To find the agent with the highest valuation, one common alternative is to hold an auction (Krishna 2009) under some protocols such as the well-known Vickrey-Clarke- Groves (VCG) mechanism (Vickrey 1961; Clarke 1971; Groves 1973). However, the payments under VCG will all be delivered to the auctioneer, which againsts our nonprofit purpose.

Entity Coreference Resolution is the task of resolving all the mentions in a document that refer to the same real world entity and is considered as one of the most difficult tasks in natural language understanding. While in it is not an end task, it has been proved to improve downstream natural language processing tasks such as entity linking, machine translation, summarization and chatbots. We conducted a systematic a review of neural-based approached and provide a detailed appraisal of the datasets and evaluation metrics in the field. Emphasis is given on Pronoun Resolution, a subtask of Coreference Resolution, which has seen various improvements in the recent years. We conclude the study by highlight the lack of agreed upon standards and propose a way to expand the task even further.

We propose a novel paradigm of semi-supervised learning (SSL)--the semi-supervised multiple representation behavior learning (SSMRBL). SSMRBL aims to tackle the difficulty of learning a grammar for natural language parsing where the data are natural language texts and the 'labels' for marking data are parsing trees and/or grammar rule pieces. We call such 'labels' as compound structured labels which require a hard work for training. SSMRBL is an incremental learning process that can learn more than one representation, which is an appropriate solution for dealing with the scarce of labeled training data in the age of big data and with the heavy workload of learning compound structured labels. We also present a typical example of SSMRBL, regarding behavior learning in form of a grammatical approach towards domain-based multiple text summarization (DBMTS). DBMTS works under the framework of rhetorical structure theory (RST). SSMRBL includes two representations: text embedding (for representing information contained in the texts) and grammar model (for representing parsing as a behavior). The first representation was learned as embedded digital vectors called impacts in a low dimensional space. The grammar model was learned in an iterative way. Then an automatic domain-oriented multi-text summarization approach was proposed based on the two representations discussed above. Experimental results on large-scale Chinese dataset SogouCA indicate that the proposed method brings a good performance even if only few labeled texts are used for training with respect to our defined automated metrics.

Successfully navigating a complex environment to obtain a desired outcome is a difficult task, that up to recently was believed to be capable only by humans. This perception has been broken down over time, especially with the introduction of deep reinforcement learning, which has greatly increased the difficulty of tasks that can be automated. However, for traditional reinforcement learning agents this requires an environment to be able to provide frequent extrinsic rewards, which are not known or accessible for many real-world environments. This project aims to explore and contrast existing reinforcement learning solutions that circumnavigate the difficulties of an environment that provide sparse rewards. Different reinforcement solutions will be implemented over a several video game environments with varying difficulty and varying frequency of rewards, as to properly investigate the applicability of these solutions. This project introduces a novel reinforcement learning solution, by combining aspects of two existing state of the art sparse reward solutions.

Recently, neural networks have shown promising results on Document-level Aspect Sentiment Classification (DASC). However, these approaches often offer little transparency w.r.t. their inner working mechanisms and lack interpretability. In this paper, to simulating the steps of analyzing aspect sentiment in a document by human beings, we propose a new Hierarchical Reinforcement Learning (HRL) approach to DASC. This approach incorporates clause selection and word selection strategies to tackle the data noise problem in the task of DASC. First, a high-level policy is proposed to select aspect-relevant clauses and discard noisy clauses. Then, a low-level policy is proposed to select sentiment-relevant words and discard noisy words inside the selected clauses. Finally, a sentiment rating predictor is designed to provide reward signals to guide both clause and word selection. Experimental results demonstrate the impressive effectiveness of the proposed approach to DASC over the state-of-the-art baselines.

Deep Reinforcement Learning (Deep RL) has been receiving increasingly more attention thanks to its encouraging performance on a variety of control tasks. Yet, conventional regularization techniques in training neural networks (e.g., $L_2$ regularization, dropout) have been largely ignored in RL methods, possibly because agents are typically trained and evaluated in the same environment. In this work, we present the first comprehensive study of regularization techniques with multiple policy optimization algorithms on continuous control tasks. Interestingly, we find conventional regularization techniques on the policy networks can often bring large improvement on the task performance, and the improvement is typically more significant when the task is more difficult. We also compare with the widely used entropy regularization and find $L_2$ regularization is generally better. Our findings are further confirmed to be robust against the choice of training hyperparameters. We also study the effects of regularizing different components and find that only regularizing the policy network is typically enough. We hope our study provides guidance for future practices in regularizing policy optimization algorithms.