Projects
Embedded Files
While I try to update this regularly (no longer true!), my Kaggle profile does a better job of compiling my latest projects and articles.
While I try to update this regularly (no longer true!), my Kaggle profile does a better job of compiling my latest projects and articles.
Projects are open for all to contribute, so "Fork, Build, and Explore Freely"
Projects are open for all to contribute, so "Fork, Build, and Explore Freely"
Focal loss is a modification of cross-entropy loss designed to address class imbalance in machine learning. It down-weights the loss assigned to well-classified examples, focusing more on hard-to-classify instances.
Problem Statement: Credit Card Fraud Identification
Have you ever wondered why we exclusively employ the sigmoid function to transform linear regression results into logistic regression classes? Why do we rely solely on the sigmoid function when there are numerous other functions, such as MinMaxScaler and Standard Scaler, that can achieve the same goal (bring values between 0 and 1 or an interval)?
18 Stars and 6 Forks
• Reverse-engineered Adobe’s 2012 ML model to identify malware in Windows Executable files (.exe), resulting in a roughly 1.5% accuracy increase (likely statistical noise).
• Enhanced the model’s explainability using LIME (Local Interpretable Model-agnostic Explanations), enabling developers to identify the root cause of a 6.7% false positive rate, which wasn’t possible in 2012.
Page updated
Report abuse