Artificial intelligence

Scientific journal

ISSN 2710-1673

ONLINE: ISSN 2710-1681

Select your language


Ефективне збільшення області притягнення глобального мінімуму квадратичного бінарного функціонала при нейромережному пошуку

Kryzhanovsky B.1, Karandashev J.1
1 Center for Optical and Neural Technologies NIISI RAS

Full text (PDF)

UDC: 004.8:004.9
Publication Language: Russian
Stuc. intelekt. 2009; 14(4):37-44

Abstract: A quadratic binary functional minimization problem is considered. To effectively increase the deep minima domains of attraction it is suggested to raise a matrix which constructed the functional on to some power, and to solve the minimization problem on the new obtained functional. By the example of matrixes of the twodimensional Ising’s model it is shown in the paper that suggested technique leads to a shift of local minima spectrum towards the region of deeper minima, reduces sharply the number of minima found, and gives an opportunity to find the global minimum with a probability on 3 – 4 orders greater.

Keywords:

References:

View full text (PDF)