On Machine Learning Knowledge Representation In The Form Of Partially Unitary Operator. Knowledge Generalizing Operator
A new form of ML knowledge representation with high generalization power is developed and implemented numerically. Initial πΌπ attributes and πππ class label are transformed into the corresponding Hilbert spaces by considering localized wavefunctions. A partially unitary operator optimally converting a state from πΌπ Hilbert space into πππ Hilbert space is then built from an optimization problem of transferring maximal possible probability from πΌπ to πππ, this leads to the formulation of a new algebraic problem. Constructed Knowledge Generalizing Operator π° can be considered as a πΌπ to πππ quantum channel; it is a partially unitary rectangular matrix of the dimension dim(πππ) Γdim(πΌπ) transforming operators as A^πππ=π° A^πΌππ°^β . Whereas only operator π° projections squared are observable β¨πππ|π°|πΌπβ©^2 (probabilities), the fundamental equation is formulated for the operator π° itself. This is the reason of high generalizing power of the approach; the situation is the same as for the SchrΓΆdinger equation: we can only measure Ο^2, but the equation is written for Ο itself.
READ FULL TEXT