Tinn (Tiny Neural Network) is a 200 line dependency free neural network library written in C99.
For a demo on how to learn hand written digits, get some training data:
wget http://archive.ics.uci.edu/ml/machine-learning-databases/semeion/semeion.data
make; ./test
The training data consists of hand written digits written both slowly and quickly. Each line in the data set corresponds to one handwritten digit. Each digit is 16x16 pixels in size giving 256 inputs to the neural network.
At the end of the line 10 digits signify the hand written digit:
0: 1 0 0 0 0 0 0 0 0 0
1: 0 1 0 0 0 0 0 0 0 0
2: 0 0 1 0 0 0 0 0 0 0
3: 0 0 0 1 0 0 0 0 0 0
4: 0 0 0 0 1 0 0 0 0 0
...
9: 0 0 0 0 0 0 0 0 0 1
This gives 10 outputs to the neural network. The test program will output the accuracy for each digit. Expect above 99% accuracy for the correct digit, and less that 0.1% accuracy for the other digits.
Portable - Runs where a C99 or C++98 compiler is present.
Sigmoidal activation.
One hidden layer.
Tinn will never use more than the C standard library.
Tinn is great for embedded systems. Train a model on your powerful desktop and load it onto a microcontroller and use the analog to digital converter to predict real time events.
The Tinn source code will always be less than 200 lines. Functions externed in the Tinn header are protected with the xt namespace standing for externed tinn.
Tinn can easily be multi-threaded with a bit of ingenuity but the master branch will remain single threaded to aid development for embedded systems.
Tinn does not seed the random number generator. Do not forget to do so yourself.
Always shuffle your input data. Shuffle again after every training iteration.
Get greater training accuracy by annealing your learning rate. For instance, multiply your learning rate by 0.99 every training iteration. This will zero in on a good learning minima.
Tinn is a practice in minimalism.
Tinn is not a fully featured neural network C library like Kann, or Genann:
https://github.com/attractivechaos/kann
https://github.com/codeplea/genann
Rust: https://github.com/dvdplm/rustinn
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。
1. 开源生态
2. 协作、人、软件
3. 评估模型