Nastya Kan

Bayesian statistics is a way of thinking about probability that helps us make decisions and predictions by combining what we already know (called a prior) with new data we see. It is a large field with many popular applications (bayesian networks, diffusion models, variational autoencoders), with a couple key... See more

Follow for more @unfoldedai Let’s say we have a dataset with 2 classes. If our data is not linearly separable, a simple linear model like logistic regression or a linear SVM will struggle to find an accurate decision boundary. However, neural networks provide a powerful alternative by... See more

When I applied for my first roles in Data Science, I didn’t have a technical degree just a solid story told through my resume. Here’s what actually made the difference👇 💡 1. Hackathons I joined a few beginner-friendly hackathons (Kaggle, Companies, university events). Even if... See more

Your complete guide to the most powerful ML algorithms in 2025! Whether you’re a beginner or an experienced data scientist, these 10 algorithms will level up your machine learning game: - Random Forest — The reliable ensemble champion - XGBoost — Kaggle competition... See more

Follow for more @getintoai Neural networks are used to approximate functions when we have inputs and want to generate specific outputs without knowing the exact function (we need an approximation). We do this by training the network on examples, providing it with pairs of inputs and... See more

The dot product is a math operation that takes two equal length vectors and produces a scalar (which is a single value). Numerically, it is calculated by multiplying corresponding components of the vectors and summing the results. Graphically, the dot product is the product of the magnitudes... See more

🌟 Master Data Science in 2025 – Your Complete 11-Month Roadmap! 🌟 If you’ve been dreaming about becoming a Data Scientist but don’t know where to start, this video is your ultimate guide. In just 11 months, you can go from beginner to job-ready by following this step-by-step learning path. 📅... See more

Follow for more @unfoldedai In a neural network, each layer performs math computations to transform the input data into a final output. At the start, input data is passed through the first layer, where it’s multiplied by weights and added to biases, followed by an activation function that introduces some... See more

In 2023, Meta intern Guangxuan Xiao discovered that removing the first few tokens in a sliding-window KV cache caused catastrophic degradation in long-context LLM performance. These tokens acted as attention sinks, stabilizing attention distributions due to softmax’s requirement that weights sum to one. The simple fix—pinning the first four... See more