Joanikij Chulev's Projects
This project explores the capabilities of machine learning in identifying musical instruments from audio samples. It leverages the rich NSynth dataset, featuring a diverse collection of instrument sounds. Utilizing a diverse range of Advanced Machine Learning and DL methods, this project aims to classify both form numerical and image data.
This project integrates wheel, ultrasonic, infrared, bump, and Bluetooth sensors into an Arduino-coded autonomous vehicle. Challenges include vehicle engineering and coding for accurate sensing. Testing validates sensor functionality and coding efficacy. The vehicle must navigate a maze, complete turns and was made to act completely autonomously.
A simple Java program designed to interact with users through a character-based menu interface. This program offers a user-friendly way to navigate various options and functionalities using character inputs. It serves as a practical demonstration of implementing menu-driven functionality and char mutations in Java.
This project simulates the college network environment in 2014. By replicating network structures and traffic patterns, it aims to offer insights into the network's performance, security, and scalability during that time period.
This project involves building multiple deep reinforcement learning agents to interact with gymnasium environments. Exercise 1 focuses on constructing a DQN agent for solving the CartPole-v1 environment. Exercise 2 extends to building a DQN agent for any similar environment like LunarLander-v2. Exercise 3 involves constructing a modern DDQN agent.
This project conducts an experimental investigation into the packets of Gmail and Outlook. By analyzing network traffic data, it aims to compare the packet characteristics and protocol usage between the two email services.
This assignment delves into image processing tasks using OpenCV in Python. Tasks include histogram operations, value changes, applying smoothing filters, removing noise with median filters, and sharpening via unsharp masking. Practical implementation showcases OpenCV's versatility and effectiveness in addressing image processing challenges.
This exploration delves into graph theory, focusing on Minimum Spanning Trees (MSTs) and algorithms like Prim's and Kruskal's. Using a geographical graph of Zeeland towns, real-life distances as edge weights are employed. The study analyzes execution times, visualizes MSTs, and applies Dijkstra's algorithm for shortest paths.
This project focuses on analyzing packet size distribution and protocol breakdown for various websites and various packet formats. Through examining packet data in various formats it aims to uncover insights into network traffic patterns, helping understand data transmission behaviors and optimize network performance.
This study conducts a comprehensive analysis of sorting algorithms, evaluating their performance in execution time. Three experiments were conducted: comparing algorithms on randomly generated data, assessing parallel sorting algorithms on large datasets, and introducing my novel sorting algorithm with a simple naming convention - chulevsort.
A recommendation system developed in Prolog, designed to assist users in making fast food choices. Leveraging expert system techniques, this program utilizes a knowledge base of fast food items and their features. By considering user preferences, it provides personalized recommendations for fast food options that align with their needs.
This project explores a 3-link robotic finger's capabilities, focusing on precise fingertip positioning. Modeling in Mathematica, it examines workspace, simplifies to 2D space, and develops an iterative inverse kinematics solver for accurate joint angle determination, addressing real-world constraints.
We explore the Fourier Transform's significance in converting signals from time to frequency domains, focusing on Discrete FT (DFT) for digital images. DFT coefficients' magnitude and phase offer insights into signal characteristics, aiding signal processing and image analysis. Fast Fourier Transform (FFT) methods streamline computation.
This experiment assesses the influence of learning rates on Gradient Descent convergence time across three mathematical functions. With five learning rates per function, convergence times were averaged and compared using both numerical and symbolic differentiation implementations.
The project involves merging three schemas into a comprehensive database for a Mars habitat. By integrating schemas related to habitat infrastructure, resource management, and personnel records, this project aims to create a unified schema that facilitates efficient data management and retrieval.
This project focuses on cleansing medical SQL data to ensure accuracy and reliability. Through rigorous data cleaning techniques, including handling missing values, removing duplicates, standardizing formats, and identifying outliers, this project aims to prepare the medical dataset for further analysis.
This paper shows in detail an approach to signal and image reconditioning to a certain extent where the image or signal were deemed to be restorable. We aimed to represent how normalized convolution would work to return images to their original state.
This report explores sampling methods for discrete and continuous probability distributions. It explores various methods for obtaining samples and it also investigates techniques for transforming samples from one distribution to another, showcasing their versatility in handling diverse probability density functions.
This project contains the source code and documentation for an advanced image processing project focused on the detection of prohibition road signs. The project leverages OpenCV techniques to identify and recognize signs characterized by their distinctive red annuli. 3 pipelines are tested and implemented.
Relationship-Logic-Table is a Java program structured with multiple classes to analyze relationships stored in text files. By parsing input data from text files, this program employs logical operations to determine the relationship status and info between individuals. It offers a systematic approach to understanding and visualizing connections.
This report focuses on understanding the Kawasaki RS03N robotic arm. Part I involves programming the arm using its teach pendant, culminating in a creative construction task with wooden Kapla blocks. The project aimed for intricate block positioning, settling on building a sitting figure with angled arms.
This project showcases the diverse properties of Fourier, Z, and Laplace transforms using Mathematica. Through interactive demonstrations and visualizations, it illustrates the transformational capabilities of each technique. The aim is providing a deeper understanding of signal processing and system analysis methodologies.
We explore the classic Mice & Cheese problem using threads. By simulating mice attempting to reach cheese in a maze concurrently, it delves into multi-threading concepts such as synchronization, deadlock avoidance, and resource management. The aim is implementing concurrent algorithms and understanding thread interactions in a simulated space.
This project tackles US death causes misrepresentation by merging data from Google Trends, US CDC, and NYT. By understanding the data sources, applying data wrangling techniques, and employing visualization methods, it aims to reveal accurate insights into US mortality, addressing existing research gaps.
This project employs Multiple Linear Regression to model average college faculty salaries in the US. By analyzing various factors, it aims to provide insights into the determinants of faculty compensation, aiding in understanding salary trends and informing policy decisions in higher education.