Gustav Louw

No, assuming that efopen() was successful, fclose() and fprintf() can be used without worry.

I rely on runtime analysis with valgrind. Static analysis is typically overkill at times. The metrics supplied by gcc's -Wall and -Wextra is usually enough.

Developments done for the most part as this is a practice in minimalism, If you're looking for a fully developed neural net library then checkout FANN (https://github.com/libfann/fann).

It may be simpler to separate the other training data into 90% train and 10% test. One can do so with head and tail in unix.

Yup, try this: https://pastebin.com/raw/9L7UV7wD

The iris data set uses floating point inputs and string outputs. > > 5.0,3.3,1.4,0.2, Iris-setosa > ... > 5.7,2.8,4.1,1.3, Iris-versicolor > ... > 6.3,2.9,5.6,1.8, Iris-virginica Try remapping these strings to...

Thanks for checking back

Do as you like, its MIT. I can link your wrapper in the readme

I think the neural network situation on rust is simlar - only heavy set libraries exist. A straight port of Tinn to Rust would be very impressive and useful in...

Thats correct. just xtsave() it and xtload() it next time. You can use xtpredict() all you like with the loaded network. I do so in an embedded setting where I...