Category: AI/ML Research

Machine learning models have become increasingly sophisticated, but this complexity often comes at the cost of interpretability.

Quantization is a frequently used strategy applied to production machine learning models, particularly large and complex ones, to make them lightweight by reducing the numerical precision of the model’s parameters (weights) — usually from 32-bit floating-point to lower representations like 8-bit integers.

This post is divided into five parts; they are: • Naive Tokenization • Stemming and Lemmatization • Byte-Pair Encoding (BPE) • WordPiece • SentencePiece and Unigram The simplest form of tokenization splits text into tokens based on whitespace.

Machine learning model development often feels like navigating a maze, exciting but filled with twists, dead ends, and time sinks.

In machine learning model development, feature engineering plays a crucial role since real-world data often comes with noise, missing values, skewed distributions, and even inconsistent formats.

Learning machine learning can be challenging.

This article is divided into three parts; they are: • Full Transformer Models: Encoder-Decoder Architecture • Encoder-Only Models • Decoder-Only Models The original transformer architecture, introduced in “Attention is All You Need,” combines an encoder and decoder specifically designed for sequence-to-sequence (seq2seq) tasks like machine translation.

“I’m feeling blue today” versus “I painted the fence blue.

We have seen a new era of agentic IDEs like Windsurf and Cursor AI.

If you’ve been into machine learning for a while, you’ve probably noticed that the same books get recommended over and over again.