Estimation Of False Recognition Rate Of Face Recognition Offline Learning Algorithm Integrals And Commutators

Age invariant face recognition(AIFR)is highly required in many applications.Even for humans,identifying faces of different ages has certain difficulties.Therefore,it poses a unique challenge to computer vision systems.There are many mature deep learning methods for face recognition,such as Multi-task Cascaded Convolutional Networks.However,in some special application scenarios,the operation efficiency of deep learning algorithms is relatively low,which cannot meet the needs of practical applications.Probabilistic dynamic programming is an algorithmic idea that uses state recursion to solve probability or expectations.Probabilistic dynamic programming also finds the probability of the entire event by finding the probability of a series of sub-events of an event.When using dynamic programming to solve a problem,how to define the state of a problem is the key to solving this problem and directly determines the efficiency of the algorithm.This paper proposes an offline learning algorithm that does not participate in neural networks to adapt to changes in face over time,and uses probabilistic dynamic programming to estimate the effect of offline algorithms on the misrecognition rate over time.The offline learning algorithm proposed in this paper does not require the use of external data sets and the introduction of external features,nor does it require time-consuming training,and can be updated in real time;The offline learning algorithm proposed in this paper is significantly faster than the AIFR method based on deep learning;This paper derives an approximate formula for the misrecognition rate of the offline learning algorithm,which makes the accuracy of the algorithm predictable.The algorithm is theoretically better interpretable than deep learning.