UC Riverside

According to its mathematical description, a Hopfield Neural Network serves as a content addressable memory with binary threshold units. The elements in that memory consist of the correlations between elements of memory vectors. In this thesis, the

feasibility of a Hopfield Neural Network using DNA molecules as the working substance is introduced. In addition, I present an experimental study proving that forming a DNA based memory storing the information of two different 6-bit black and white images,

representing memory vectors, and recalling one of original images with the use of a partial image are possible. It is observed that the recalling with a DNA based Hopfield Neural Network using incomplete inputs is more powerful comparing to theoretical one

using corrupted inputs. Moreover, as a supplementary work, I show that application of T4 Gene 32 Protein to Isothermal Linear Amplification (ILA) reduces the production of fragment DNA strands, one of the biggest problems of this type of amplification.