In this paper, we address two key challenges in deep reinforcement learning setting, sample inefficiency and slow learning, with a dual NN-driven learning approach. In the proposed approach, we use two deep NNs with independent initialization to robustly approximate the action-value function in the presence of image inputs. In particular, we develop a temporal difference (TD) error-driven learning approach, where we introduce a set of linear transformations of the TD error to directly update the parameters of each layer in the deep NN. We demonstrate theoretically that the cost minimized by the error-driven learning (EDL) regime is an approximation of the empirical cost and the approximation error reduces as learning progresses, irrespective of the size of the network. Using simulation analysis, we show that the proposed methods enables faster learning and convergence and requires reduced buffer size (thereby increasing the sample efficiency).

Neural Architecture Search (NAS) methods have been successfully applied to image tasks with excellent results. However, NAS methods are often complex and tend to converge to local minima as soon as generated architectures seem to yield good results. In this paper, we propose G-EA, a novel approach for guided evolutionary NAS. The rationale behind G-EA, is to explore the search space by generating and evaluating several architectures in each generation at initialization stage using a zero-proxy estimator, where only the highest-scoring network is trained and kept for the next generation. This evaluation at initialization stage allows continuous extraction of knowledge from the search space without increasing computation, thus allowing the search to be efficiently guided. Moreover, G-EA forces exploitation of the most performant networks by descendant generation while at the same time forcing exploration by parent mutation and by favouring younger architectures to the detriment of older ones. Experimental results demonstrate the effectiveness of the proposed method, showing that G-EA achieves state-of-the-art results in NAS-Bench-201 search space in CIFAR-10, CIFAR-100 and ImageNet16-120, with mean accuracies of 93.98%, 72.12% and 45.94% respectively.

Such process is called the pruning of Lightweight model design has become an important direction the neural network. The pruned neural network can usually in the application of deep learning technology, obtain much faster inference speed than the original model, pruning is an effective mean to achieve a large reduction in which has high significance in actual deployment when the model parameters and FLOPs. The existing neural network efficiency of the model is critical. The pruning of neural network pruning methods mostly start from the importance of parameters, can usually be seen as a three-step pipeline: training and design parameter evaluation metrics to perform the original model, parameter pruning, and fine-tuning the parameter pruning iteratively. These methods are not pruned model. Thus most of these network pruning methods studied from the perspective of model topology, may be effective are data-related, i.e, when the model training is completed, but not efficient, and requires completely different the parameters are pruned according to their values, pruning for different datasets. In this paper, we study the which means that for different datasets, the pruned neural graph structure of the neural network, and propose regular network is different; while some pruning methods perform graph based pruning (RGP) to perform a one-shot neural pruning on the initialized model, such as the lottery ticket network pruning. We generate a regular graph, set the hypothesis [7], but it is still implemented by pruning after node degree value of the graph to meet the pruning ratio, pre-training the original model, it uses the initialized parameters and reduce the average shortest path length of the graph by to reset the pruned model so that an initialized subnetwork swapping the edges to obtain the optimal edge distribution.

Deep Reinforcement Learning (RL) has emerged as a powerful paradigm to solve a range of complex yet specific control tasks. Yet training generalist agents that can quickly adapt to new tasks remains an outstanding challenge. Recent advances in unsupervised RL have shown that pre-training RL agents with self-supervised intrinsic rewards can result in efficient adaptation. However, these algorithms have been hard to compare and develop due to the lack of a unified benchmark. To this end, we introduce the Unsupervised Reinforcement Learning Benchmark (URLB). URLB consists of two phases: reward-free pre-training and downstream task adaptation with extrinsic rewards. Building on the DeepMind Control Suite, we provide twelve continuous control tasks from three domains for evaluation and open-source code for eight leading unsupervised RL methods. We find that the implemented baselines make progress but are not able to solve URLB and propose directions for future research.

Data packet routing in aeronautical ad-hoc networks (AANETs) is challenging due to their high-dynamic topology. In this paper, we invoke deep learning (DL) to assist routing in AANETs. We set out from the single objective of minimizing the end-to-end (E2E) delay. Specifically, a deep neural network (DNN) is conceived for mapping the local geographic information observed by the forwarding node into the information required for determining the optimal next hop. The DNN is trained by exploiting the regular mobility pattern of commercial passenger airplanes from historical flight data. After training, the DNN is stored by each airplane for assisting their routing decisions during flight relying solely on local geographic information. Furthermore, we extend the DL-aided routing algorithm to a multi-objective scenario, where we aim for simultaneously minimizing the delay, maximizing the path capacity, and maximizing the path lifetime. Our simulation results based on real flight data show that the proposed DL-aided routing outperforms existing position-based routing protocols in terms of its E2E delay, path capacity as well as path lifetime, and it is capable of approaching the Pareto front that is obtained using global link information.

Current maritime communications mainly rely on satellites having meager transmission resources, hence suffering from poorer performance than modern terrestrial wireless networks. With the growth of transcontinental air traffic, the promising concept of aeronautical ad hoc networking relying on commercial passenger airplanes is potentially capable of enhancing satellite-based maritime communications via air-to-ground and multi-hop air-to-air links. In this article, we conceive space-air-ground integrated networks (SAGINs) for supporting ubiquitous maritime communications, where the low-earth-orbit satellite constellations, passenger airplanes, terrestrial base stations, ships, respectively, serve as the space-, air-, ground- and sea-layer. To meet heterogeneous service requirements, and accommodate the time-varying and self-organizing nature of SAGINs, we propose a deep learning (DL) aided multi-objective routing algorithm, which exploits the quasi-predictable network topology and operates in a distributed manner. Our simulation results based on real satellite, flight, and shipping data in the North Atlantic region show that the integrated network enhances the coverage quality by reducing the end-to-end (E2E) delay and by boosting the E2E throughput as well as improving the path-lifetime. The results demonstrate that our DL-aided multi-objective routing algorithm is capable of achieving near Pareto-optimal performance.

Recently, interactive recommendation systems based on reinforcement learning have been attended by researchers due to the consider recommendation procedure as a dynamic process and update the recommendation model based on immediate user feedback, which is neglected in traditional methods. The existing works have two significant drawbacks. Firstly, inefficient ranking function to produce the Top-N recommendation list. Secondly, focusing on recommendation accuracy and inattention to other evaluation metrics such as diversity. This paper proposes a deep reinforcement learning based recommendation system by utilizing Actor-Critic architecture to model dynamic users' interaction with the recommender agent and maximize the expected long-term reward. Furthermore, we propose utilizing Spotify's ANNoy algorithm to find the most similar items to generated action by actor-network. After that, the Total Diversity Effect Ranking algorithm is used to generate the recommendations concerning relevancy and diversity. Moreover, we apply positional encoding to compute representations of the user's interaction sequence without using sequence-aligned recurrent neural networks. Extensive experiments on the MovieLens dataset demonstrate that our proposed model is able to generate a diverse while relevance recommendation list based on the user's preferences.

We propose the molecular omics network (MOOMIN) a multimodal graph neural network that can predict the synergistic effect of drug combinations for cancer treatment. Our model captures the representation based on the context of drugs at multiple scales based on a drug-protein interaction network and metadata. Structural properties of the compounds and proteins are encoded to create vertex features for a message-passing scheme that operates on the bipartite interaction graph. Propagated messages form multi-resolution drug representations which we utilized to create drug pair descriptors. By conditioning the drug combination representations on the cancer cell type we define a synergy scoring function that can inductively score unseen pairs of drugs. Experimental results on the synergy scoring task demonstrate that MOOMIN outperforms state-of-the-art graph fingerprinting, proximity preserving node embedding, and existing deep learning approaches. Further results establish that the predictive performance of our model is robust to hyperparameter changes. We demonstrate that the model makes high-quality predictions over a wide range of cancer cell line tissues, out-of-sample predictions can be validated with external synergy databases, and that the proposed model is data-efficient at learning.

Deep learning frameworks such as TensorFlow and PyTorch provide a productive interface for expressing and training a deep neural network (DNN) model on a single device or using data parallelism. Still, they may not be flexible or efficient enough in training emerging large models on distributed devices, which require more sophisticated parallelism beyond data parallelism. Plugins or wrappers have been developed to strengthen these frameworks for model or pipeline parallelism, but they complicate the usage and implementation of distributed deep learning. Aiming at a simple, neat redesign of distributed deep learning frameworks for various parallelism paradigms, we present OneFlow, a novel distributed training framework based on an SBP (split, broadcast and partial-value) abstraction and the actor model. SBP enables much easier programming of data parallelism and model parallelism than existing frameworks, and the actor model provides a succinct runtime mechanism to manage the complex dependencies imposed by resource constraints, data movement and computation in distributed deep learning. We demonstrate the general applicability and efficiency of OneFlow for training various large DNN models with case studies and extensive experiments. The results show that OneFlow outperforms many well-known customized libraries built on top of the state-of-the-art frameworks. The code of OneFlow is available at: https://github.com/Oneflow-Inc/oneflow.

Pedestrian trajectory prediction is a key technology in many applications such as video surveillance, social robot navigation, and autonomous driving, and significant progress has been made in this research topic. However, there remain two limitations of previous studies. First, with the continuation of time, the prediction error at each time step increases significantly, causing the final displacement error to be impossible to ignore. Second, the prediction results of multiple pedestrians might be impractical in the prediction horizon, i.e., the predicted trajectories might collide with each other. To overcome these limitations, this work proposes a novel trajectory prediction method called CSR, which consists of a cascaded conditional variational autoencoder (CVAE) module and a socially-aware regression module. The cascaded CVAE module first estimates the future trajectories in a sequential pattern. Specifically, each CVAE concatenates the past trajectories and the predicted points so far as the input and predicts the location at the following time step. Then, the socially-aware regression module generates offsets from the estimated future trajectories to produce the socially compliant final predictions, which are more reasonable and accurate results than the estimated trajectories. Moreover, considering the large model parameters of the cascaded CVAE module, a slide CVAE module is further exploited to improve the model efficiency using one shared CVAE, in a slidable manner. Experiments results demonstrate that the proposed method exhibits improvements over state-of-the-art method on the Stanford Drone Dataset (SDD) and ETH/UCY of approximately 38.0% and 22.2%, respectively.