Sapere Aude. "Dare to know." — Horace
温故知新 (Onko Chishin) "Cherishing old knowledge, acquiring new." — Japanese Idiom
I am a 4th-year Computer Engineering student at Manisa Celal Bayar University and an active Working Group Member of the European COST Action CA22145 (GameTable). My academic focus is defined by a specific pursuit: bridging the "Epistemic Gap" between high-performance Deep Learning models and human cognitive understanding. I operate under the handle Aidoneus—a reference to the "Unseen"—reflecting my goal to make the hidden logic of "Black Box" algorithms visible, explainable, and pedagogically valuable.
My primary area of interest lies at the intersection of Cognitive Game AI, Explainable AI (XAI), and Human-Computer Interaction (HCI). I investigate how neural networks process strategic decisions in traditional board games (specifically Go/Weiqi) and how these machine intuitions can be translated into human-readable concepts (Chunking).
Key Methodologies:
-
Axiomatic Attribution: Utilizing Integrated Gradients (IG) and Grad-CAM to visualize neural network rationale, deliberately moving beyond unstable methods like LIME.
-
Deep Learning & Computer Vision: Developing Convolutional Neural Networks (ResNet architectures) and Object Detection models (YOLO) for board state evaluation and digitization.
-
Cognitive Alignment: Translating statistical AI outputs into "Interaction Primitives" to foster human cognitive skill acquisition.
Status: Graduation Project I (Final Report Published)
A "Glass Box" Deep Learning framework designed to open the black box of Go AI engines. This project utilizes a ResNet-18 model trained on 150,000+ professional games and implements Integrated Gradients to visualize the strategic rationale behind expert move generation.
Live Demo: Streamlit Page
2. Beyond the Move: Phase II (In Progress)
Status: Pre-Graduate Research / IEEE CoG Preparation
Transitioning from pixel-level attribution to high-level "Intuition Modeling." Upgrading the visual pipeline with YOLOv8 for robust physical board detection and implementing Grad-CAM to capture human-like "Chunking" and shape recognition (e.g., Atsumi).
I am open to academic collaboration and discussion regarding AI ethics and game strategy.
- ResearchGate: İpek Naz Sipahi
- Academic Mail: My mail
- LinkedIn: İpek Naz Sipahi
- Institution: Manisa Celal Bayar University, Türkiye
"The true sign of intelligence is not knowledge but imagination."