Research On Emotional Quotient Network Model And Algorithm

With the development of big data,Internet of Things(IoT)and cloud computing technology,the accumulated massive data and enhanced computing capabilities have enabled artificial intelligence(AI)to get rid of the limits of data scale and computational power.AI has achieved unprecedented success in many fields such as computer vision,natural language processing and autonomous driving.However,with the increase in people's demand for more advanced AI,the traditional "intelligence quotient(IQ)" type of AI has gradually exposed shortcomings such as poor open scene compatibility,weak multi-intelligence collaboration and lack of perceptual thinking.Therefore,how to realize a strong"Emotional Quotient(EQ)" type of AI has become an important research content in the future.At this stage,research on emotional intelligence still faces many challenges.First,it is difficult to perform effective intelligent collaboration based on traditional intelligent nodes,and the lack of networking mechanisms makes it difficult to guarantee the feasibility of multi-agent networks.Second,the"instantaneous-process duality" feature of emotional data leads to failures of existing data processing approaches;Third,the current centralized-training and single-model intelligent paradigm is difficult to effectively perform high-dimensional and personalized affective computing;Finally,the black-box AI model is too uncontrollable due to the lack of emotional dynamics,which results in that it is difficult to guarantee the stability.In view of the above problems,this thesis focuses on the model and algorithm of emotional intelligence network.Firstly,a kind of swarm intelligence architecture inspired by human society network is proposed.Based on this architecture,we carry out in-depth research on the making of system macro rules,emotional memory processing,affective computing and emotional control of the emotional intelligence network with the use of random pulse theory,memory level neural network,hidden Markov model and van der Pol oscillator theory.The main work and the results are summarized as follows:(1)This thesis proposes a systematic emotion random pulse model.The prediction of systematic emotions is realized based on the transfer of basic emotions.By adding pulse signals,the systematic emotion is controlled in a stable operating space.Based on this model,an algorithm for determining the pulse rules is proposed to realize the formulation of macro rules of the emotional intelligence network.Through persistent and global attractiveness analysis,it is proved that the emotional intelligence network running under this rule can achieve effective feasibility guarantee.(2)A kind of memory level neural network is proposed.The memory transition mechanism implemented by the continuous memory switch enables the memory level neuron to automatically convert between the instantaneous and time series data of emotions.We network the memory level neurons and propose its learning algorithm.The network trained by this algorithm can realize accurate analysis and prediction of long-term emotional memory.(3)This thesis proposes an amygdala-inspired affective computing framework,which realizes machine affective computing by observing the representation status of emotional intelligence nodes.First,the convolutional neural network is compressed based on pruning and hashing tricks to achieve accelerated recognition of external emergency emotions.Then,a deeper memory level neural network is realized by adding the time pooling layer to achieve accurate recognition of external process emotions.Finally,based on the recognition results of external emotions,a personalized Markov model of intracranial emotions is established.(4)A Van der Pol model of emotional dynamics was proposed.First,the restoring force of emotional fluctuations is nonlinearized.Then,the model is solved and the steady-state response analysis is performed respectively in the autonomous and random-excitation cases.According to the orbit of emotional state,four health states of emotional system are defined.Based on the dynamic model,a stable feedback control algorithm for emotions of EQ nodes is proposed.