Heyo,
just published Version 0.3 of my own C++ deep learning framework. It includes a variety of optimization Algorithms like Nesterov Momentum, RMSProp, Adam etc. Furthermore I've done some changes to the architecture giving more space for custom settings. Thank you for my amazing journey through arcade 🙂.
repo: github.com/Neurologism/brainet
Hey yall,
I'm releasing the next version of my Project. I'm writing my own deep learning framework in C++ since 3 months now.
The release contains some new features like bagging and dropout. However it's mainly about major bugfixes and a rewrite of the model interface.
Feel free to reach out to me if you want to learn more about deep learning frameworks.🙂
github.com/Neurologism/brainet
Hey yall,
today I'm publishing Version 0.2.0 of my own deep learning library called Brainet. It adds various regularization techniques, bugfixes and improves Optimization in comparison with Version 0.1.0. Feel free to contact me if you want to learn more about building a deep learning framework. If you want to try it out, the README is a good place to start. Happy about any feedback 🙂.
See ya
github.com/Neurologism/brainet/tree/v0.2.0
Heyo,
I finally did. I finished version 0.1 of my project. For more than 2 months I was working on my own deep learning framework. It's entirely written in C++ and only based on the STL, so everyone with a C++ compiler should be able to run it. It's nothing incredible impressive so far, TensorFlow is much better, but I'm just getting started. 😛 That's mostly because a solid foundation for a big framework it's quite hard. I think I've rewritten every single line of code at least 3 times.
The current UI allows a user to build a fully connected neural network. You can configure the number of layers, the neurons in a layer, the activation function of each layer and so on. The only supported dataset is currently mnist as I haven't found time to write readers for other input formats. An Example program can be found in test.cpp. Neither convolutional layers or recurrent /recursive layers are currently supported, but will be added in Version 0.3. Version 0.2 will be adding a range of regularization and optimization algorithms. So stay tuned 🙂.
Feel free to fork the repo, play with the UI or the code and feel absolutely free to contact me via Slack or Email: [email protected] if you have questions, suggestions or just want to learn how to build your own deep learning library. Till next time 😮
Note: only a small percentage of the invested time was logged in arcade, the more than 500 commits probably give a better impression
github.com/Neurologism/brainet
