Writing
My thoughts and explorations on AI, Machine Learning, Deep Learning and Data Science topics.
- Gradient Descent (Sep 2026)
- LangChain (Sep 2026)
- Beyond the Basics, Exploring RobustScaler, MaxAbsScaler and Other Scaling Methods (Dec 2024)
- StandardScaler vs MinMaxScaler (Dec 2024)
- How Was My First Steps in Machine Learning: Lessons and Surprises (Oct 2024)
- A Journey into the World of Soroush and Beyond! (Sep 2024)
A self-contained crash course on how learning actually happens: the optimization problem, batch / stochastic / mini-batch GD, why vanilla GD fails, momentum and first moment, second moment and adaptive rates (AdaGrad, RMSProp), Adam with bias correction, AdamW and modern schedules — with full mathematics, convergence rates and a practical training playbook.
Read the full version · Crash Course · 11 lessons
A self-contained crash course on LangChain 1.x for Python — models, prompts, LCEL, structured output, state and memory, retrieval and RAG, tools, agents, streaming, observability, evaluation and production concerns. Each lesson builds on the last with copy-ready code and primary sources.
Read the full version · Crash Course · 15 lessons
In this article, I dive into advanced scaling
techniques that go beyond traditional methods like StandardScaler and MinMaxScaler. The article
highlights RobustScaler, which uses median and interquartile
range to handle outliers effectively, and
MaxAbsScaler, perfect for sparse data as it preserves zero
entries. It also explores additional methods
like QuantileTransformer and PowerTransformer, which offer solutions for non-linear scaling and
variance
stabilization. This comprehensive guide helps data scientists choose the right scaler for complex
datasets, boosting model accuracy and robustness.
Full version of the Article · Published on DataDrivenInvestor
Data scaling is a key preprocessing step in machine learning that transforms data into a consistent
range, boosting model performance and interpretability. The article highlights two widely used
techniques: StandardScaler and MinMaxScaler. StandardScaler adjusts data to have a mean of 0 and a
standard deviation of 1, making it perfect for normally distributed data by eliminating bias and
ensuring balance. Conversely, MinMaxScaler scales data to a fixed range (typically 0 to 1), maintaining
proportional relationships and excelling with bounded or non-normal data.
Full version of the Article · Published on Towards AI
As someone with a background in industrial engineering, I had always been attracted by solving complex
problems and searching for the way to enhance the efficiency of systems. When I, about a year ago for
the first time, started exploring machine learning, I knew it was a powerful tool with enormous
potential, though at that time, I did not realize fully how it would expand my perspective. It has been
an exciting journey, challenging and full of unexpected lessons. In this post, I'll share a bit
about
what I've learned and some surprises that came along the way.
Full version of the Blog
Hello, and welcome to my World!
I'm so glad you're here. This blog marks the start of something I've wanted to create for
a long time—a space to connect, share, and explore. In a way, this is my online home, a place where I
can introduce myself and give you a glimpse of what am I doing.
Full version of the Blog