Andreas Karatzas's Projects
Hosting my home pc configuration
(AAAI 2018) Action Branching Architectures for Deep Reinforcement Learning
[ECCV 2018] AMC: AutoML for Model Compression and Acceleration on Mobile Devices
My personal repository.
Linux/OSX/FreeBSD resource monitor
BISMO: A Scalable Bit-Serial Matrix Multiplication Overlay for Reconfigurable Computing
Simulator for BitFusion
Implementation of Block Recurrent Transformer - Pytorch
Brevitas: quantization-aware training in PyTorch
This is a Combinational Circuit Logic Simulation Tool. There is a C++ version and a C version.
Add a command-line interface to any C++ program
Modern cryptography algorithms using deep learning
In this repository, there is an implementation of CSV to Graph<V, E> transform using Python 3.8 . The repository was created to validate the results of the TSP algorithm using a naive approach, which is implemented here.
Simple neural network implementation using CUDA technology. It is an educational implementation.
Curiosity-driven Exploration by Self-supervised Prediction
Towards autonomous drones using deep learning.
Red wine quality prediction based on multi-dimensional vectors. Each dimension is a different sensor metric. In the same repository, there is another model that predicts if a news title is fake or not (onion-or-not dataset).
Implementation of DDPG agent in PyTorch.
Movie recommendation using deep learning techniques. There are two different implementations. The first is based on fully convolutional networks, and the second utilizes the efficiency of embeddings. The dataset of the project was the IMDB dataset.
Neural Network Distiller by Intel AI Lab: a Python package for neural network compression research. https://intellabs.github.io/distiller
Crawler for big data analysis.
Faster R-CNN fully customizable implementation using PyTorch 1.10.
This is an implementation of a genetic algorithm for educational purposes. The implementation was done from scratch using plain NumPy and CuPy with Python 3.8. The purpose of the algorithm was to recommend movies to users.
A utility for stressing GPUs by driving utilization (and thus power consumption) up and down in user-defined cycle intervals. It will also randomly drop power consumption down to idle and spike it back up
Custom graphics driver using Verilog on Xilinx FPGA platform.
Implementation of DDPG with Hindsight Experience Replay agent in PyTorch.
Client - Server offloading of DNNs.
Control the HP Omen keyboard lighting and performance settings in Linux
Easily turn large sets of image urls to an image dataset. Can download, resize and package 100M urls in 20h on one machine.