Juan Raúl Padrón Griffe

About me
I am a postdoctoral researcher in Junqiu Zhu's group at Shandong University. My research focuses on photorealistic rendering, including appearance modeling for complex multi-scale materials (human skin, hair, fur, scales, feathers, cosmetics, textiles, and granular materials), light transport simulation algorithms and data-driven representations. I am also interested in inverse rendering, differentiable rendering and 3D reconstruction. I am currently collaborating with Zahra Montazeri's group on the representation and rendering of textiles.

I completed my PhD dissertation, Modeling and Rendering of Multiscale Materials, at the Graphics and Imaging Lab, Universidad de Zaragoza. I was a Marie Sklodowska-Curie Fellow of the EU Project PRIME, advised by Adolfo Muñoz and Adrian Jarabo. I hold an M.Sc. in Informatics from the Technical University of Munich with a major in Graphics and Vision, and a B.Sc. in Computer Science from the Universidad Central de Venezuela with a major in Computer Graphics.

Projects

Deep Reinforcement Learning for Protein Folding

2018, Oct 02    

The protein structure determines properties and functions of the protein. Therefore, scientistic can develop drugs for this specific unique protein shape in order to cure diseases. Nowadays, Machine Learning methods arise as an alternative to the costly experimentation techniques (cryo-electron microscopy, nuclear magnetic resonance, X-ray crystallography) to help accelerate research.

For the Hands-on Deep Learning for Computer Vision practical course at the Technical University of Munich, we implement the infamous AlphaGo algorithm in C++ to address the protein folding problem with the reinforcement learning approach. The main idea of the algorithm is training a neural network to estimate the policy and the values estimates, where the policy is improved by looking ahead into the future via Monte Carlo Tree Search guided by the value network. The environment (Rosetta 2017.39) and the Monte Carlo Tree Search algorithm were implemented in C++ 17, while the neural network was written in Python (TensorFlow 1.4.1). We rely on the TensorRT library to improve significantly the performance of the neural network inference. As a result, we could speed up the implementation 18.95 times regarding to the python version. Nevertheless, the approach is still unfeasible for large proteins (e.g. L = 332).

Results overfitting a single protein:

Smooth Return Value Loss Policy Loss

Team Members: Juan Raul Padron Griffe, Matthias Humt, Felix Opolka
Instructors: Vladimir Golkov, Daniel Cremers

Presentation

DeepMind implemented successfully this project idea in a system called AlphaFold. If you want to know more about cool projects in biomedicine, please contact Vladimir Golkov. Felix Opolka, which was the main author of this project, is currently working on cool publications about neural networks for graph-structured data.

Twitter
# Protein Folding # Reinforcement Learning