Ten years into the revival of deep networks and artificial intelligence, we propose a theoretical framework that sheds light on understanding deep networks within a bigger picture of intelligence in general. We introduce two fundamental principles, Parsimony and Self-consistency, which address two fundamental questions regarding intelligence: what to learn and how to learn, respectively. We believe the two principles serve as the cornerstone for the emergence of intelligence, artificial or natural. While they have rich classical roots, we argue that they can be stated anew in entirely measurable and computable ways. More specifically, the two principles lead to an effective and efficient computational framework, compressive closed-loop transcription, which unifies and explains the evolution of modern deep networks and most practices of artificial intelligence. While we use mainly visual data modeling as an example, we believe the two principles will unify understanding of broad families of autonomous intelligent systems and provide a framework for understanding the brain.
Visual recognition of cardiac images is important for cardiac pathology diagnosis and treatment. Due to the limited availability of annotated datasets, traditional methods usually extract features directly from two-dimensional slices of three-dimensional (3D) heart images, followed by pathological classification. This process may not ensure the overall anatomical consistency in 3D heart. A new method for classification of cardiac pathology is therefore proposed based on 3D parametric model reconstruction. First, 3D heart models are reconstructed based on multiple 3D volumes of cardiac imaging data at the end-systole (ES) and end-diastole (ED) phases. Next, based on these reconstructed 3D hearts, 3D parametric models are constructed through the statistical shape model (SSM), and then the heart data are augmented via the variation in shape parameters of one 3D parametric model with visual knowledge constraints. Finally, shape and motion features of 3D heart models across two phases are extracted to classify cardiac pathology. Comprehensive experiments on the automated cardiac diagnosis challenge (ACDC) dataset of the Statistical Atlases and Computational Modelling of the Heart (STACOM) workshop confirm the superior performance and efficiency of this proposed approach.
In fifth-generation wireless communication system (5G), more connections are built between metaheuristics and electromagnetic equipment design. In this paper, we propose a self-adaptive grey wolf optimizer (SAGWO) combined with a novel optimization model of a 5G frequency selection surface (FSS) based on FSS unit nodes. SAGWO includes three improvement strategies, improving the initial distribution, increasing the randomness, and enhancing the local search, to accelerate the convergence and effectively avoid local optima. In benchmark tests, the proposed optimizer performs better than the five other optimization algorithms: original grey wolf optimizer (GWO), genetic algorithm (GA), particle swarm optimizer (PSO), improved grey wolf optimizer (IGWO), and selective opposition based grey wolf optimization (SOGWO). Due to its global searchability, SAGWO is suitable for solving the optimization problem of a 5G FSS that has a large design space. The combination of SAGWO and the new FSS optimization model can automatically obtain the shape of the FSS unit with electromagnetic interference shielding capability at the center operating frequency. To verify the performance of the proposed method, a double-layer ring FSS is designed with the purpose of providing electromagnetic interference shielding features at 28 GHz. The results show that the optimized FSS has better electromagnetic interference shielding at the center frequency and has higher angular stability. Finally, a sample of the optimized FSS is fabricated and tested.
Smart meters in the Power Internet of Things generate a large amount of power data. However, data privacy in the process of calculation, storage, and transmission is an urgent problem to be solved. Therefore, in this paper we propose a power resource dispatching framework (PRDF) with a privacy protection function, which uses a certificateless aggregate signcryption scheme based on cloud-fog cooperation. Using pseudonyms and aggregating users’ power data, PRDF not only protects users’ privacy, but also reduces the computing cost and communication overhead under traditional cloud computing. In addition, if the control center finds that a user has submitted abnormal data, it can send a request to the user management center to track the real identity of the user. Our scheme satisfies security requirements based on the random oracle model, including confidentiality and unforgeability. Furthermore, we compare our scheme with other certificateless aggregate signcryption schemes by simulations. Simulation results show that compared with traditional methods, our method performs better in terms of the computation cost.
Network coding can save wireless network resources and is very fast in comparison with traditional routing. In real application scenarios, network coding is vulnerable to pollution attacks and forgery attacks. To solve these problems, the certificateless broadcast multi-signature for network coding (NC-CLBMS) method is devised, where each source node user generates a multi-signature about the message vector, and the intermediate node linearly combines the received data. NC-CLBMS is a multi-source multi-signature method with anti-pollution and anti-forgery advantages; moreover, it has a fixed signature length and its computation efficiency is very high. NC-CLBMS has extensive application prospects in unmanned aerial vehicle (UAV) communication networks, fifth-generation wireless networks, wireless sensor networks, mobile wireless networks, and Internet of Vehicles.
Line-of-sight (LoS) probability prediction is critical to the performance optimization of wireless communication systems. However, it is challenging to predict the LoS probability of air-to-ground (A2G) communication scenarios, because the altitude of unmanned aerial vehicles (UAVs) or other aircraft varies from dozens of meters to several kilometers. This paper presents an altitude-dependent empirical LoS probability model for A2G scenarios. Before estimating the model parameters, we design a K-nearest neighbor (KNN) based strategy to classify LoS and non-LoS (NLoS) paths. Then, a two-layer back propagation neural network (BPNN) based parameter estimation method is developed to build the relationship between every model parameter and the UAV altitude. Simulation results show that the results obtained using our proposed model has good consistency with the ray tracing (RT) data, the measurement data, and the results obtained using the standard models. Our model can also provide wider applicable altitudes than other LoS probability models, and thus can be applied to different altitudes under various A2G scenarios.
We propose a competitive binary multi-objective grey wolf optimizer (CBMOGWO) to reduce the heavy computational burden of conventional multi-objective antenna topology optimization problems. This method introduces a population competition mechanism to reduce the burden of electromagnetic (EM) simulation and achieve appropriate fitness values. Furthermore, we introduce a function of cosine oscillation to improve the linear convergence factor of the original binary multi-objective grey wolf optimizer (BMOGWO) to achieve a good balance between exploration and exploitation. Then, the optimization performance of CBMOGWO is verified on 12 standard multi-objective test problems (MOTPs) and four multi-objective knapsack problems (MOKPs) by comparison with the original BMOGWO and the traditional binary multi-objective particle swarm optimization (BMOPSO). Finally, the effectiveness of our method in reducing the computational cost is validated by an example of a compact high-isolation dual-band multiple-input multiple-output (MIMO) antenna with high-dimensional mixed design variables and multiple objectives. The experimental results show that CBMOGWO reduces nearly half of the computational cost compared with traditional methods, which indicates that our method is highly efficient for complex antenna topology optimization problems. It provides new ideas for exploring new and unexpected antenna structures based on multi-objective evolutionary algorithms (MOEAs) in a flexible and efficient manner.
A functional neuron has been developed from a simple neural circuit by incorporating a phototube and a thermistor in different branch circuits. The physical field energy is controlled by the photocurrent across the phototube and the channel current across the thermistor. The firing mode of this neuron is controlled synchronously by external temperature and illumination. There is energy diversity when two functional neurons are exposed to different illumination and temperature conditions. As a result, synapse connections can be created and activated in an adaptive way when field energy is exchanged between neurons. We propose two kinds of criteria to discuss the enhancement of synapse connections to neurons. The energy diversity between neurons determines the increase of the coupling intensity and synaptic current for neurons, and the realization of synchronization is helpful in maintaining energy balance between neurons. The first criterion is similar to the saturation gain scheme in that the coupling intensity is increased with a constant step within a certain period until it reaches energy balance or complete synchronization. The second criterion is that the coupling intensity increases exponentially before reaching energy balance. When two neurons become non-identical, phase synchronization can be controlled during the activation of synapse connections to neurons. For two identical neurons, the second criterion for taming synaptic intensity is effective for reaching complete synchronization and energy balance, even in the presence of noise. This indicates that a synapse connection may prefer to enhance its coupling intensity exponentially. These results are helpful in discovering why synapses are awaken and synaptic current becomes time-varying when any neurons are excited by external stimuli. The potential biophysical mechanism is that energy balance is broken and then synapse connections are activated to maintain an adaptive energy balance between the neurons. These results provide guidance for designing and training intelligent neural networks by taming the coupling channels with gradient energy distribution.