Learning to Search for Job Shop Scheduling via Deep Reinforcement Learning

Recent studies in using deep reinforcement learning (DRL) to solve Job-shop scheduling problems (JSSP) focus on construction heuristics. However, their performance is still far from optimality, mainly because the underlying graph representation scheme is unsuitable for modeling partial solutions at each construction step. This paper proposes a novel DRL-based method to learn improvement heuristics for JSSP, where graph representation is employed to encode complete solutions. We design a Graph Neural Network based representation scheme, consisting of two modules to effectively capture the information of dynamic topology and different types of nodes in graphs encountered during the improvement process. To speed up solution evaluation during improvement, we design a novel message-passing mechanism that can evaluate multiple solutions simultaneously. Extensive experiments on classic benchmarks show that the improvement policy learned by our method outperforms state-of-the-art DRL-based methods by a large margin.

Decision-making with Imaginary Opponent Models

Opponent modeling has benefited a controlled agent's decision-making by constructing models of other agents. Existing methods commonly assume access to opponents' observations and actions, which is infeasible when opponents' behaviors are unobservable or hard to obtain. We propose a novel multi-agent distributional actor-critic algorithm to achieve imaginary opponent modeling with purely local information (i.e., the controlled agent's observations, actions, and rewards). Specifically, the actor maintains a speculated belief of the opponents, which we call the \textit{imaginary opponent models}, to predict opponents' actions using local observations and makes decisions accordingly. Further, the distributional critic models the return distribution of the policy. It reflects the quality of the actor and thus can guide the training of the imaginary opponent model that the actor relies on. Extensive experiments confirm that our method successfully models opponents' behaviors without their data and delivers superior performance against baseline methods with a faster convergence speed.

Bootstrapping meaning through listening: Unsupervised learning of spoken sentence embeddings

Inducing semantic representations directly from speech signals is a highly challenging task but has many useful applications in speech mining and spoken language understanding. This study tackles the unsupervised learning of semantic representations for spoken utterances. Through converting speech signals into hidden units generated from acoustic unit discovery, we propose WavEmbed, a multimodal sequential autoencoder that predicts hidden units from a dense representation of speech. Secondly, we also propose S-HuBERT to induce meaning through knowledge distillation, in which a sentence embedding model is first trained on hidden units and passes its knowledge to a speech encoder through contrastive learning. The best performing model achieves a moderate correlation (0.5~0.6) with human judgments, without relying on any labels or transcriptions. Furthermore, these models can also be easily extended to leverage textual transcriptions of speech to learn much better speech embeddings that are strongly correlated with human annotations. Our proposed methods are applicable to the development of purely data-driven systems for speech mining, indexing and search.

Learning to Solve Multiple-TSP with Time Window and Rejections via Deep Reinforcement Learning

We propose a manager-worker framework based on deep reinforcement learning to tackle a hard yet nontrivial variant of Travelling Salesman Problem (TSP), \ie~multiple-vehicle TSP with time window and rejections (mTSPTWR), where customers who cannot be served before the deadline are subject to rejections. Particularly, in the proposed framework, a manager agent learns to divide mTSPTWR into sub-routing tasks by assigning customers to each vehicle via a Graph Isomorphism Network (GIN) based policy network. A worker agent learns to solve sub-routing tasks by minimizing the cost in terms of both tour length and rejection rate for each vehicle, the maximum of which is then fed back to the manager agent to learn better assignments. Experimental results demonstrate that the proposed framework outperforms strong baselines in terms of higher solution quality and shorter computation time. More importantly, the trained agents also achieve competitive performance for solving unseen larger instances.

Multi-scale Sampling and Aggregation Network For High Dynamic Range Imaging

High dynamic range (HDR) imaging is a fundamental problem in image processing, which aims to generate well-exposed images, even in the presence of varying illumination in the scenes. In recent years, multi-exposure fusion methods have achieved remarkable results, which merge multiple low dynamic range (LDR) images, captured with different exposures, to generate corresponding HDR images. However, synthesizing HDR images in dynamic scenes is still challenging and in high demand. There are two challenges in producing HDR images: 1). Object motion between LDR images can easily cause undesirable ghosting artifacts in the generated results. 2). Under and overexposed regions often contain distorted image content, because of insufficient compensation for these regions in the merging stage. In this paper, we propose a multi-scale sampling and aggregation network for HDR imaging in dynamic scenes. To effectively alleviate the problems caused by small and large motions, our method implicitly aligns LDR images by sampling and aggregating high-correspondence features in a coarse-to-fine manner. Furthermore, we propose a densely connected network based on discrete wavelet transform for performance improvement, which decomposes the input into several non-overlapping frequency subbands and adaptively performs compensation in the wavelet domain. Experiments show that our proposed method can achieve state-of-the-art performances under diverse scenes, compared to other promising HDR imaging methods. In addition, the HDR images generated by our method contain cleaner and more detailed content, with fewer distortions, leading to better visual quality.

Sampling Efficient Deep Reinforcement Learning through Preference-Guided Stochastic Exploration

Massive practical works addressed by Deep Q-network (DQN) algorithm have indicated that stochastic policy, despite its simplicity, is the most frequently used exploration approach. However, most existing stochastic exploration approaches either explore new actions heuristically regardless of Q-values or inevitably introduce bias into the learning process to couple the sampling with Q-values. In this paper, we propose a novel preference-guided $\epsilon$-greedy exploration algorithm that can efficiently learn the action distribution in line with the landscape of Q-values for DQN without introducing additional bias. Specifically, we design a dual architecture consisting of two branches, one of which is a copy of DQN, namely the Q-branch. The other branch, which we call the preference branch, learns the action preference that the DQN implicit follows. We theoretically prove that the policy improvement theorem holds for the preference-guided $\epsilon$-greedy policy and experimentally show that the inferred action preference distribution aligns with the landscape of corresponding Q-values. Consequently, preference-guided $\epsilon$-greedy exploration motivates the DQN agent to take diverse actions, i.e., actions with larger Q-values can be sampled more frequently whereas actions with smaller Q-values still have a chance to be explored, thus encouraging the exploration. We assess the proposed method with four well-known DQN variants in nine different environments. Extensive results confirm the superiority of our proposed method in terms of performance and convergence speed. Index Terms- Preference-guided exploration, stochastic policy, data efficiency, deep reinforcement learning, deep Q-learning.

Applying Feature Underspecified Lexicon Phonological Features in Multilingual Text-to-Speech

This study investigates whether the phonological features derived from the Featurally Underspecified Lexicon model can be applied in text-to-speech systems to generate native and non-native speech in English and Mandarin. We present a mapping of ARPABET/pinyin to SAMPA/SAMPA-SC and then to phonological features. This mapping was tested for whether it could lead to the successful generation of native, non-native, and code-switched speech in the two languages. We ran two experiments, one with a small dataset and one with a larger dataset. The results supported that phonological features could be used as a feasible input system for languages in or not in the train data, although further investigation is needed to improve model performance. The results lend support to FUL by presenting successfully synthesised output, and by having the output carrying a source-language accent when synthesising a language not in the training data. The TTS process stimulated human second language acquisition process and thus also confirm FUL's ability to account for acquisition.

ByT5 model for massively multilingual grapheme-to-phoneme conversion

In this study, we tackle massively multilingual grapheme-to-phoneme conversion through implementing G2P models based on ByT5. We have curated a G2P dataset from various sources that covers around 100 languages and trained large-scale multilingual G2P models based on ByT5. We found that ByT5 operating on byte-level inputs significantly outperformed the token-based mT5 model in terms of multilingual G2P. Pairwise comparison with monolingual models in these languages suggests that multilingual ByT5 models generally lower the phone error rate by jointly learning from a variety of languages. The pretrained model can further benefit low resource G2P through zero-shot prediction on unseen languages or provides pretrained weights for finetuning, which helps the model converge to a lower phone error rate than randomly initialized weights. To facilitate future research on multilingual G2P, we make available our code and pretrained multilingual G2P models at: https://github.com/lingjzhu/CharsiuG2P.

Spherical Convolution empowered FoV Prediction in 360-degree Video Multicast with Limited FoV Feedback

Field of view (FoV) prediction is critical in 360-degree video multicast, which is a key component of the emerging Virtual Reality (VR) and Augmented Reality (AR) applications. Most of the current prediction methods combining saliency detection and FoV information neither take into account that the distortion of projected 360-degree videos can invalidate the weight sharing of traditional convolutional networks, nor do they adequately consider the difficulty of obtaining complete multi-user FoV information, which degrades the prediction performance. This paper proposes a spherical convolution-empowered FoV prediction method, which is a multi-source prediction framework combining salient features extracted from 360-degree video with limited FoV feedback information. A spherical convolution neural network (CNN) is used instead of a traditional two-dimensional CNN to eliminate the problem of weight sharing failure caused by video projection distortion. Specifically, salient spatial-temporal features are extracted through a spherical convolution-based saliency detection model, after which the limited feedback FoV information is represented as a time-series model based on a spherical convolution-empowered gated recurrent unit network. Finally, the extracted salient video features are combined to predict future user FoVs. The experimental results show that the performance of the proposed method is better than other prediction methods.