jie.hang's Projects
An adversarial example library for constructing attacks, building defenses, and benchmarking both
Fine-tune CNN in Keras
Source code examples from the Parallel Forall Blog
we provide the design scheme of the course assistance learning system on android platform on the basis of exploring the key technology related to the development of android and SQLite database, and successfully exploit a software called “The Course Assistance Learning System” based on the Android 6.0 version eventually. Due to the people’s operating habits and particular demands, the course table is made to expand the application to the user’s mobile terminal equipment, and to provide auxiliary functions about some course information management. Therefore, students can easily view the course at any time, download resources, and further improve the efficiency of students’ learning.
Sample codes for my CUDA programming book
CUDA official sample codes
Source code that accompanies The CUDA Handbook.
cudahandbook中的代码以及书籍
cudnn samples are provided to verify if the operator's calculation results are correct on other gpu products(e.g., NPU, TPU) compatible to NVIDIA cuda/cudnn.
Software that can generate photos from paintings, turn horses into zebras, perform style transfer, and more.
Convert D3M raw dataset to D3M clean dataset with Featuretools
A library containing both highly optimized building blocks and an execution engine for data pre-processing in deep learning applications
Dask tutorial
Keras code and weights files for popular deep learning models.
A simple and accurate method to fool deep neural networks
Transparent Cudnn / Cublas / Eigen usage for the deep learning training using MNIST dataset.
Traffic Sign Recognition - Fine tuning VGG16 + GTSRB
MIT课程《Distributed Systems 》学习和翻译
Deep learning for time-varying multi-entity datasets
A toolkit for making real world machine learning and data analysis applications in C++
Apache Spark docker image
The reader is design and development on the Android platform, according to the analysis of the mobile electronic reader customer demand analysis and technical feasibility, through the stage of the overall design and detailed design coding testing and electronic reader software design. This software design is not only to achieve the e-book reader the most basic reading function, but also for the user to provide the support function of the rich, on the phone to achieve local book search, add books, books show, books and delete functions. Reading books is mainly to provide users with some of the functions of reading, such as flip effects, the font size of the regulation, the brightness of the regulation, reading schedule adjustment, catalog view. In the process of design and development, through continuous testing, continuous improvement, the function of the mobile phone electronic reader gradually stabilized. Finally, the software has been tested and verified, basically able to meet the needs of users.
Experiment codes for GMBA
TensorFlow Code for paper "Efficient Neural Architecture Search via Parameter Sharing"
PyTorch implementation of "Efficient Neural Architecture Search via Parameters Sharing"
Integrated a feature squeezing layer into a defensive distillation system to evade adversarial attacks. It reduced the effectiveness of adversarial attacks from 100% to 11%.
Circumventing the defense in "Ensemble Adversarial Training: Attacks and Defenses"
Ensemble Adversarial Training on MNIST
Random Forests and Multi-class AdaBoost for Wine and MNIST datasets