Hermite Tensor
A symmetric tensor formed by the coefficients of a higher-order expansion of a function in the Hermite polynomial basis. It is commonly used to analyze function learning problems under Gaussian distributions.
Timeline (last 90 days)
The algorithm uses implicit tensor computation techniques to build a low-order filtering procedure based on the Hermite tensor of the objective function, in order to amplify the separation between weight vectors.
The Hermite tensor is the symmetric tensor formed by the high-order expansion coefficients of a function in the Hermite polynomial basis, and is commonly used to analyze function learning problems under the Gaussian distribution.
For a positive linear combination of ReLUs of the form f(x)=Σ_i a_i σ(w_i^T x), the low-order part of its Hermite tensor implicitly contains the direction information of the weight vectors w_i.
All Facts (3)
The algorithm uses implicit tensor computation techniques to build a low-order filtering procedure based on the Hermite tensor of the objective function, in order to amplify the separation between weight vectors.
50%UnverifiedThe Hermite tensor is the symmetric tensor formed by the high-order expansion coefficients of a function in the Hermite polynomial basis, and is commonly used to analyze function learning problems under the Gaussian distribution.
50%UnverifiedFor a positive linear combination of ReLUs of the form f(x)=Σ_i a_i σ(w_i^T x), the low-order part of its Hermite tensor implicitly contains the direction information of the weight vectors w_i.
50%