Arrays that were near singular are more likely to be noticed now, and a lot of linear algebra routines give slightly different results now (in some cases big differences due to the way that round-off can accumulate badly when working with linear algebra.) If you happen to be in a situation where you need bit-for-bit compatability with older runs, then you could struggle a lot with the newer version. The linear algebra libraries were upgraded multiple times since R2013a.GPUs that were supported in the R2013a time are generally not supported now, with more advanced GPUs being required.If you were doing simple NN work before, it probably would not be all that much work to rewrite. By R2013a, some of the older routines were already recommended against, but the older routines have been completely removed now. There has been a lot of evolution in Neural Network facilities.In most cases it should not be difficult to rewrite. In particular, SVM has changed interfaces since then. However, Statistics includes "Machine Learning" these days, which has had more and different interfaces added.some of the statistical routines involving distributions changed a bit since R2013a, but not a lot.In some cases, making adaptations could be a lot of work. What is now called the "Legacy" tools were 32 bit only, and the interface was redesigned for 64 bit use.
Data Acquistion Toolbox changed a lot between R2013a and current releases.
0 kommentar(er)
