Mahdi Molavi | Senior Backend Engineer & Tensor Researcher https://mahdimolavi.ir/ en Mahdi Molavi | Senior Backend Engineer & Tensor Researcher DataLife Engine Backend Notes 5: From Tensor Research to Backend Engineering https://mahdimolavi.ir/backend-notes/31-backend-notes-5-from-tensor-research-to-backend-engineering.html https://mahdimolavi.ir/backend-notes/31-backend-notes-5-from-tensor-research-to-backend-engineering.html Backend Notes 5: From Tensor Research to Backend Engineering

A personal note on what carried over from a master's thesis on tensor decompositions to more than ten years of .NET backend work: the shape of data, updating instead of recomputing, honest measurement and the math behind tools.

]]> admin Mon, 02 Nov 2026 10:00:00 +0330 Tensor Notes 5: From Brain Scans to Streaming Data https://mahdimolavi.ir/tensor-decompose/21-tensor-notes-5-brain-scans-streaming-data.html https://mahdimolavi.ir/tensor-decompose/21-tensor-notes-5-brain-scans-streaming-data.html Tensor Notes 5: From Brain Scans to Streaming Data

Detecting early mild cognitive impairment against healthy controls in resting-state fMRI with tc-SVD, and updating a tensor decomposition as new data streams in instead of recomputing it.

]]> admin Thu, 29 Oct 2026 10:00:00 +0330 IRT Calibration Notes 6: Equal Sample Sizes Are Not Equating https://mahdimolavi.ir/backend-notes/81-irt-calibration-notes-6-equal-sample-sizes-are-not-equating.html https://mahdimolavi.ir/backend-notes/81-irt-calibration-notes-6-equal-sample-sizes-are-not-equating.html IRT Calibration Notes 6: Equal Sample Sizes Are Not Equating

Equal N per track made year-to-year IRT comparisons more controlled, once a subtopic-key bug that hid 13 of 30 questions was fixed. Bar charts, boxplots, item information ranks and the scale-linking problem show why equal sample sizes are still not equating.

]]> admin Wed, 28 Oct 2026 10:00:00 +0330 IRT Calibration Notes 5: Twenty Samples per Size in a Reproducible Sampling Study https://mahdimolavi.ir/backend-notes/79-irt-calibration-notes-5-reproducible-repeated-sampling-study.html https://mahdimolavi.ir/backend-notes/79-irt-calibration-notes-5-reproducible-repeated-sampling-study.html IRT Calibration Notes 5: Twenty Samples per Size in a Reproducible Sampling Study

A nine-exam repeated-sampling study of IRT discrimination, built to be reproducible and restartable. It covers nested samples from deterministic seeds, fit keys that expose reused fits, atomic checkpoints and paired tests with Benjamini-Hochberg control.

]]> admin Tue, 27 Oct 2026 10:00:00 +0330