Name: Christian Adriano, PhD candidate
Type: User
Company: Hasso-Plattner Institut
Bio: Bayesian Learning, Causal Inference, Software Architecture. https://www.linkedin.com/in/christian-medeiros-adriano-65813a/
Location: Potsdam, Germany
Blog: https://christianadriano.com/
Christian Adriano, PhD candidate's Projects
hackathon!!!
Classification of subjects professions using Bayesian Random Forest and Bayesian Logistic Regression
Scripts created to compute posterior probabilities
A repo for holding example code
Explore causal models for the accuracy of fault understanding
Computes different complexity metrics, Halstead, LOCs, Cyclomatic Complexity, etc.
Investigating approaches to prevent confounding in deep networks
Infrastructure to crowdsource software debugging by means of a question and answer approach.
Experimentation with crowd debugging via questions and answers
Utility functions and examples of data wrangling
Tests for stationarity using R and Python libraries
Descriptive statistics based on graphs metrics
Data wrangling scripts to preprocess data from crowdsourced microtasks
Processing various datasets from Self-driving accidents and crowdsourcing tasks
Investigating a Large Language Model (LLM) based tool for explaning software redesign
Investigating a Large Language Model (LLM) based tool for explaning bug fixes
Expectation Maximization to build Gaussian mixture models of the duration of task executions.
data visualization for crowd fault localization data
Project to study how utility theory to match workers to answers with the goal minimizing the number of questions asked in order to locate faults. This is my first attempt to model the problems of when to stop asking questions as a sequential decision problem.
Basic algorithms I have been playing with
Markov Chain model to prioritize tasks (fault localization questions to ask a crowd of programmers))
Explore different designs of Multi-Armed Bandits, mostly Contextual Bayesian Bandits
Daily coding practice, programming tips, best practices, etc.
Data for running the experiments
Comparative experiments with various Bayesian learning methods
Exploring Machine Learning Control for Dynamical Nonlinear Systems
Replication of experiments, Descriptive statistics (ANOVA, Chi2square, Wilcoxon, Power calculations), Correlation Analysis (Kendall-tau), Predictive models (logistic regression). Goal is to investigate the factors that impact the accuracy of fault understanding. The analyzed factors are attributes of programmers (profession, year of experience) and tasks (duration, confidence).
Select features by doing correlation analysis, PCA, decision trees, and information entropy