A huge number of information is available nowadays and it keeps increasing every day. As such, the need for recommender system to recommend relevant items or information is in high demand. Furthermore, recommender system is expected to deliver relevant items from a trustable source. In this paper, we proposed a new trust calculation that is incorporated into a hybrid recommender system. Our new trust calculation is calculated based on user's score in a particular system, and its potential implementation is demonstrated through a prototype design.

Algorithmic Differentiation (AD) is applied to numerical simulation programs in order to produce code for the computation of various kinds of derivatives. A data flow analysis called To-Be-Recorded (TBR) analysis aims to detect the set of variables required in reverse section and values that can be lost because of the overwriting in forward section for adjoint mode of AD. The implementation of the TBR analysis uses a syntax-directed algorithm based on the following attribute grammar. The expected result of TBR analysis is to get more efficient generated code in terms of memory. We have implemented TBR analysis inside our AD-tool called “Derivative Code Compiler (DCC)”. Depending on the testing cases, the percentage of memory savings can be different. In the test case illustrated in this paper, a memory saving of approximately 18% was achieved.