Writing your own neural net from scratch?


Related to: [Neural networks, machine learning, CNNs]

Hello. I am just beginning research on AI and found interests in neural networks.

I have been doing some research for about a month now, but I have searched a few things in the past. I have noticed that people like to use TensorFlow to build their AI, but I have no exposure to Python (I do C/C++). There seems to be no place to learn AI the way TensorFlow teaches it for C, so I felt that I have to do my own research and build my own net.

I have also been reading papers, but the math is too much. I know there’s no loop hole through that as the math is necessary, but I just finished 10th grade. Just got over Algebra 2…

Is is possible for me to design my own neural network and have it work? I have particular interest in CNNs. I have intermediate knowledge in C and I’m a blank slate to AI/neural network.

I was also wondering if I can “translate” my knowledge from studying how TensorFlow works (specifically the net at playground.tensorflow.org) to C. What I mean is just like defining neurons in code or something then doing the training in some way. I just started with that site, so I don’t really know much.

Also any ideas how I can implement a convnet in C? There’s lots of papers, but all they talk about is the numbers, math, and their goal, but not much about the net itself. I’ve been reading that Michael Nielsen book, and that’s the only help I’ve been getting.

*am I even on the right site to ask this??

Thanks in advance


TensorFlow is written in C++ and there is a documented C++ interface: https://www.tensorflow.org/api_guides/cc/guide