Latest News

Bayesian Neural Networks

Adapted from an appendix of my MS thesis. Bayesian Neural Network Deep neural networks (DNNs) are usually trained using a regularized maximum likeliho…

Financial Market Analysis (Double-Tower Transformer)

Content Intro - What is it? Briefly A predictive machine learning system explicitly designed to generate profitable trading signals from high-frequenc…

Transformers From Scratch In Code

map: 1) hyperparams block 2) data block loss tracking function 3) Single Attention Head Class (head_size) at first B, T, C = shape 4) head_size = n_em…

Is anyone working with the YOLO model?

GNN vs. Trees: High-Speed Hybrid Architecture for XLA Runtime Prediction

GNN vs. Trees: High-Speed Hybrid Architecture for XLA Runtime Prediction Introduction A common trap in Machine Learning engineering is deploying over-…

One "+x" That Made 100-Layer Networks Trainable: ResNet Skip Connections

Deep networks have a cruel paradox. In theory, more layers should never hurt — the extra ones could just learn to pass their input through unchanged. …

AI Deep Learning: Explained Simply

Every time someone says "AI," what they usually mean, underneath it all, is deep learning. ChatGPT, Google Photos recognizing your face, Netflix guess…

"Starting My AI/ML Journey in Public — and Backpropagation Already Broke My Brain"

Hi, I'm Chakradhar— A 3rd year BTech student trying to become AI/ML engineer right from the 1st year i had wasted all my time, but yeah whatever it is…

Transfer Learning: Stand on a Pretrained Model

You don't have a million labeled images or a GPU farm — and you don't need them. Transfer learning lets you stand on a model someone else trained and …

How Transformer Decoders Generate Text — From Causal Masking to Decoding

A Transformer Decoder does not generate a sentence all at once. It predicts one token. Then it feeds that token back and predicts the next one. That s…

GANs Explained: Two Networks That Make Fakes Real

Before diffusion models, GANs were how AI learned to generate realistic images. The idea is a game between two networks — and you can watch it reach i…

I Built the First Purely Learned Frame-by-Frame Tetris AI: Then It Started Cheating

Greetings all! You might know me from my Snake AI ablation series where I spent an unreasonable amount of time teaching a snake to eat apples. This is…

Transformers From Scratch: Assembling the Block Behind GPT

Yesterday: attention — each token deciding how much to look at every other token. Today we assemble the rest of the Transformer block around it. Stack…

Pandas and Data Visualization Using Matplotlib and Seaborn

New chapter in Learning data analytics and data science. The focus now is on Pandas as a Python library alongside Matplotlib and Seaborn for data visu…

Anthropic’s Fable/Mythos shutdown is the first real model export-control shock

Anthropic’s Fable/Mythos shutdown is the first real model export-control shock The important AI story this week is not just that Anthropic launched bi…

Tune Model Training in Real Time — Zero Latency, Zero Restarts (Kiponos Python SDK)

Training jobs are supposed to be fire-and-forget. You pick hyperparameters, launch a GPU job, and wait hours for a learning curve that tells you what …

I Built an "Amazon-Style" AI Review Summarizer for Any Dataset (NLP, Transformers, Streamlit)

Have you seen those new AI-generated review summaries on Amazon? They are incredibly useful for buyers, but there’s a catch: they are completely locke…

Loss Functions: MSE vs MAE vs Cross-Entropy, Visualized

Pick the wrong loss function and your model optimises the wrong thing — perfectly. The loss is the single number training tries to shrink, so it quiet…

Fine-Tuning Llama 3.2 3B on Medical QA: Week 4 - When Lower Loss Meant a Worse Model

What Happened This Week Week 3 produced a working fine-tuned model: one epoch, one dataset, a clear improvement over the base model. This week 4 was s…

AI vs ML vs DL — the Nested Circles That Finally Make It Click

"AI", "machine learning", "deep learning", "GenAI" — used interchangeably every day, and it's wrong. Here's the single picture that fixes it forever. …

I published two IEEE Access papers as an undergrad, one on IoT security, one on cancer detection

I'm a 6th semester CS student at COMSATS University Islamabad. Over the past few months I've been doing deep learning research alongside my coursework…

Flash Attention: what it does and why it matters

Flash Attention: what it does and why it matters Your training job is paying for an A100 at $3/hour. The loss is going down, gradients are flowing, an…

How to Become a Data Scientist in 2026

How I got here On principle, you will never catch me parading myself as a some sort of expert data scientist. Technically, that's what I do in my day …

Understanding Attention in Transformers — Intuition Before Equations

When people first hear about Transformers, they often encounter words like Query, Key, Value, and Attention Heads and feel confused. But the main idea…

PyTorch from Scratch — Part 1: Tensors, Gradients & Activations

Most people use PyTorch without really knowing what's happening underneath. This series breaks the foundations down into the simplest possible explana…

A11: A Structural Answer to AI Collapse

Modern AI models are becoming increasingly powerful, but their growing capabilities come with rising risks of degradation: the loss of rare patterns, …

NVIDIA Cosmos 3: Unifying Physical AI Reasoning and Generation with Two-Tower Architecture

Training a robot to pick up an object sounds simple until you realize how many separate systems are involved: a vision model to understand the scene, …

NVIDIA Cosmos 3: How a Two-Tower Architecture Unifies Physical AI Reasoning and Generation

Training a robot to pick up an object sounds simple until you realize how many separate systems are involved: a vision model to understand the scene, …

I distilled a 7B vision model into a 2B one for screenshots — and the 7B teacher scored worse

Code: https://github.com/P0rt/vlm-distill-screenshots Model: https://huggingface.co/p00rt/qwen2-vl-2b-screenshots-distill There's a question I keep co…

ai, deepseek, machinelearning

title: The Rise of China's LLMs: A Complete History from 2017 to 2026 published: ture description: From Wu Dao 2.0 (1.75T params) to DeepSeek V3 ($5.6…