Demystifying CNN with Mathematical Insights: A Prelude with Application to an AI-based Sustainable Solution for Diabetic Retinopathy Diagnosis
Bibliographic record
Abstract
Abstract Convolutional Neural Networks (CNNs), as a computational framework for deep learning, have gained preference over traditional machine learning techniques and simple, fully- connected neural networks due to their remarkable performance for AI applications related to images and videos, in general, computer vision. Recent advancements in CNN architectures pave the way for significant contributions across various application domains, including disease diagnosis, object detection, and image classification systems. This article elaborates on the fundamental process of CNN with mathematical insights. It provides a brief overview of the mathematical and computational intricacies that underpin CNNs. It presents streamlined derivations of key components, focusing on the pivotal mechanism of the backpropagation. A pseudo-code of the generic algorithm for CNN is presented both in component form and vectorized form, alongside an exhaustive explanation of the relevant data structures to foster comprehensive understanding. The article includes an application of classifying retinal images for diagnosing diabetic retinopathy (DR). Having started the discussion about implementation from scratch, the article sheds light on transfer learning using pre-trained models for sustained efficiency. For this, performance gains are demonstrated using state-of-the-art CNN architecture for the DR classification. This way, the article aims to equip learners, researchers, and practitioners with mathematical insights into the working of CNN for proper comprehension and stepping towards efficient model development for sustainable advancements, especially for disease diagnosis. The understanding and expertise related to CNN would add to the development of large-scale and sustainable solutions based on AI in the health sector, supporting the United Nations’ agenda 2030 for sustainability. ACM Classification Codes (ccs98): I.2.6, I.5.1, K.3.2 . MSC Codes (2020): 68T07, 68T45, 92B20
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".