About

Andrei Vladimirovich Konstantinov

This website serves as a portfolio showcasing my research and professional activities. I am a PhD in Physics and Mathematics with a strong focus on developing innovative solutions within machine learning and related fields. My background combines rigorous mathematical training with practical experience in data science and computational techniques.

Academic Background & Research Interests

I completed my undergraduate education at Peter the Great Polytechnic University, earning a degree in Applied Mathematics. My doctoral research centers around addressing challenging problems in machine learning through novel theoretical frameworks and algorithmic development. My primary research interests include:

• Hard Constraints in Neural Networks: Developing methods to enforce explicit constraints on neural network outputs, ensuring adherence to domain-specific rules and regulations.
• Hybrid Models: Decision Trees & Neural Networks: Developing novel architectures that combine the interpretability of decision trees with the representational power of neural networks, aiming to achieve both accuracy and explainability.
• Interpretable Machine Learning (xAI): Developing and applying techniques to enhance the transparency and understandability of machine learning models, enabling users to gain insights into model decision-making processes.
• Concept-Based Learning: Investigating approaches that explicitly represent and leverage human-understandable concepts within machine learning models, facilitating knowledge transfer and improving generalization capabilities.
• Multiple Instance Learning (MIL): Exploring MIL approaches for scenarios with ambiguous data labels, enabling robust learning from incomplete or uncertain information.
• Survival Analysis: Applying machine learning techniques to model time-to-event outcomes, particularly in contexts where censoring is present.
• Heterogeneous Treatment Effect: Investigating methods for identifying and modeling how treatment effects vary depending on instance features.

Publications & Theses

• Publications

  Journals	Conferences	Preprints
  ~36	~17	~7

• Theses

  • PHD (2024) Research and Development of New Interpretable Machine Learning Models based on Compositions of Weak Base Models

  • Masters degree (2020) Research and development of new Deep Gradient Boosting algorithms for solving high-dimensional problems

  • Bachelor degree (2018) Algorithm and software implementation of the level set method for solving the problem of semi-automatic segmentation of medical images

Skills & Experience

Beyond my research, I have cultivated a diverse skillset through various experiences. Participation in debate teams honed my abilities in critical thinking, persuasive communication, and active listening – skills essential for effective collaboration and presenting complex ideas clearly. Furthermore, managing a student video department provided valuable experience in creative problem-solving, storytelling, and project management within a collaborative environment. In my leisure time, I enjoy playing the guitar and piano.