Skip to content
Knowledge Distillation Explained: Logits, Softmax Temperature, and Why the Loss Is Multiplied by T² (Hinton 2015) Knowledge distillation is a way to transfer what a large model (the teacher) has learned into a smaller model (the student). The method proposed in the 2015 paper "Distilling the Knowledge in a Neural Network" by Hinton, Vinyals, and Dean is still the most basic form of it. This post walks through the three core ideas of that paper for a classifier whose output layer is a softmax, using equation.. 2026. 10. 17.
How to Copy Terminal Output to Clipboard: PowerShell Set-Clipboard, cmd clip, Mac pbcopy, Linux xclip (Copy SSH Key) When you add an SSH key to GitHub, or need to paste a long path or command output somewhere else, selecting terminal text with the mouse is easy to get wrong. One missing character or a stray line break and the key no longer works. Every operating system has a command that sends output straight to the clipboard. Add one word after your command, then paste with Ctrl+V. No right-click, no dragging.. 2026. 10. 16.
How to Delete Your GitHub Account (Step-by-Step with Screenshots) Have an old GitHub account you no longer use, or a test account you want to clean up? Deleting it takes about two minutes, but the option is tucked away at the very bottom of the settings page. Here is the whole process with screenshots. ⏱ Quick summary1. Click your profile picture (top right) → 2. Settings → 3. Account → 4. Scroll down and click Delete your account → 5. Type delete my account a.. 2026. 10. 16.
Myprotein's AI Virtual Try-On: How to Try Clothes On With One Photo (+ Fuel Coach AI Nutrition Coach) The hardest part of buying workout clothes online is the question "Will this actually look good on me?" The model photos look great, but the real thing often feels different. Myprotein Korea (myprotein.co.kr) now has a "Virtual Try-On" feature: upload a photo of yourself and AI dresses you in the item. The same site also has "Fuel Coach", an AI nutrition coach that answers questions about traini.. 2026. 10. 13.
What Is Product Hunt? A Step-by-Step Guide to Submitting Your Product If you've built a website or an app, sooner or later you ask yourself: "How do I get my first real users?" One of the best answers is Product Hunt. In this post, I'll explain what Product Hunt is and then walk you through the entire submission process, step by step, using my own project, Browser Tools, as an example.What is Product Hunt?Product Hunt is a community website where people share and .. 2026. 10. 13.
NumPy Study Part 8: Build k-means Clustering From Scratch With Only NumPy (Final Project) This is the final part of the NumPy Study series, and it is a project. We will build k-means clustering, one of the most widely used machine-learning algorithms, using only NumPy: no scikit-learn, no loops over data points.Every step uses something from an earlier part:StepNumPy ideaPartGenerate datadefault_rng, normal, concatenate1, 6Check shapesshape, reshape, newaxis2Select points in a cluste.. 2026. 10. 12.
NumPy Study Part 7: Math, Rounding, Missing Values and Linear Algebra (sqrt, round, isnan, dot, @, linalg) This part collects the numerical tools you reach for every day:Math functions: powers, roots, logs, exponentials, absolute valuesRounding: round, floor, ceil, trunc, and a surprise about .5Missing values: nan, isnan, the nan-functions, filling and droppingFloating-point comparison: why 0.1 + 0.2 == 0.3 is False, and iscloseLinear algebra: dot products, @, identity, inverse, determinant, solving .. 2026. 10. 12.
NumPy Study Part 6: Combine, Split, Sort and Count (concatenate, stack, split, sort, argsort, unique, clip) Real data rarely arrives in one tidy array. You load it in pieces, add new rows, split it into training and test sets, rank it, and count categories. This part covers the tools for all of that:Combining: concatenate, stack, vstack, hstack, column_stackAdding values: append, insert, delete (and why to use them sparingly)Splitting: split, array_splitSorting and ranking: sort, argsort, partition, t.. 2026. 10. 12.
728x90

티스토리툴바