Parameters Compressing in Deep Learning

With the popularity of deep learning tools in image decomposition and natural language processing,how to support and store a large number of parameters required by deep learning algorithms has become an urgent problem to be solved.These parameters are huge and can be as many as millions.At present,a feasible direction is to use the sparse representation technique to compress the parameter matrix to achieve the purpose of reducing parameters and reducing the storage pressure.These methods include matrix decomposition and tensor decomposition.To let vector take advance of the compressing performance of matrix decomposition and tensor decomposition,we use reshaping and unfolding to let vector be the input and output of Tensor-Factorized Neural Networks.We analyze how reshaping can get the best compress ratio.According to the relationship between the shape of tensor and the number of parameters,we get a lower bound of the number of parameters.We take some data sets to verify the lower bound.

A Privacy-Preserving Federated Learning Algorithm for Intelligent Inspection in Pumped Storage Power Station

As a distributed machine learning architecture,Federated Learning(FL)can train a global model by exchanging users’model parameters without their local data.However,with the evolution of eavesdropping techniques,attackers can infer information related to users’local data with the intercepted model parameters,resulting in privacy leakage and hindering the application of FL in smart factories.To meet the privacy protection needs of the intelligent inspection task in pumped storage power stations,in this paper we propose a novel privacy-preserving FL algorithm based on multi-key Fully Homomorphic Encryption(FHE),called MFHE-PPFL.Specifically,to reduce communication costs caused by deploying the FHE algorithm,we propose a self-adaptive threshold-based model parameter compression(SATMPC)method.It can reduce the amount of encrypted data with an adaptive thresholds-enabled user selection mechanism that only enables eligible devices to communicate with the FL server.Moreover,to protect model parameter privacy during transmission,we develop a secret sharing-based multi-key RNS-CKKS(SSMR)method that encrypts the device’s uploaded parameter increments and supports decryption in device dropout scenarios.Security analyses and simulation results show that our algorithm can prevent four typical threat models and outperforms the state-of-the-art in communication costs with guaranteed accuracy.

Velocity distribution of flow with submerged flexible vegetations based on mixing-length approach

By choosing a PVC slice to simulate flexible vegetation, we carried out experiments in an open channel with submerged flexible vegetation. A 3D acoustic Doppler velocimeter （micro ADV） was used to measure local flow velocities and Reynolds stress. The results show that hydraulic characteristics in non-vegetation and vegetation layers are totally different. In a region above the vegetation, Reynolds stress distribution is linear, and the measured velocity profile is a classical logarithmic one. Based on the concept of new-riverbed, the river compression parameter representing the impact of vegetation on river is given, and a new assumption of mixing length expression is made. The formula for time-averaged velocity derived from the expression requires less parameters and simple calculation, and is useful in applications.