## Gradient Descent, the Delta Rule and Backpropagation

If you learn about machine learning you will stumble over three terms that are
related:
Gradient descent,
the Delta rule and
backpropagation
Gradient descent is a way to find a minimum in a high-dimensional space. You
go in direction of the steepest descent.
The delta rule is an update rule for single layer perceptrons. It makes use
of gradient descent.
Backpropagation
See also
Wikipedia pages:
Gradient descent
Delta rule
Backpropagation

## On-line Handwriting Recognition of Mathematical Symbols

THIS IS JUST A DRAFT. IT WILL GET UPDATED UNTIL 15 NOVEMBER 2014! On-line handwriting recognition systems get the information how a symbol is written. In contrast, OCR only gets the pixel map. I've created a system that can be used to work with handwriting recognition systems in my bachelor's thesis. Toolkit installation The hwrt toolkit can be installed via pip: # pip install hwrt System Structure The system has a configurable PROJECT_ROOT. The configuration...

## 3D Printing

3D printing is a hot topic for quite a while now. 3D printers have become cheap enough so that many ordinary people can have 3D printers for their personal use. In this article, I would like to point out some very nice applications of 3D printing. Medicine 3D printing has basically two application in medicine: Very customized non-organic parts or organic printing. Printing a human Kidney "Magic Arms" prosthetic 3D printing gives man new face...

## interACT

interACT is an exchange program for students. I'm a student from KIT (Germany) and I went to CMU in Pittsburgh (US). The information in the following article will be most useful for students who also come from KIT and want to go to CMU. Participating universities An interACT exchange is available between the following universities: United States: CMU, Pittsburgh and Mountain View USC (Los Angeles) Germany: KIT (Karlsruhe) China: HKUST (Hong Kong) Japan Waseda University...

## One-on-One Tutoring

I have heard that the famous high quality universities in England have One-on-One tutoring. That means one tutor helps one student only. That must be incredible effective for learning. From my experience as a tutor and as a "coach" (Nachhilfelehrer) I know that there is a huge difference in One-on-One tutoring compared to One to Many tutoring. If you only have one student, you can focus on the needs of this single student. And I...

## The Twiddle Algorithm

Twiddle is an algorithm that tries to find a good choice of parameters $p$ for an algorithm $\mathcal{A}$ that returns an error. The algorithm is quite simple to implement. I guess gradient descent might be better for most cases, but Twiddle does not require any knowledge about the algorithm $\mathcal{A}$ which might be a big advantage. And you don't have to calculate the gradient of high dimensional functions, which is nice, too. The algorithm Here...

## Hyperlapse

Johannes Kopf, a researcher at Microsoft, has recently published a stunning video about a software project he seems to have participated in. The video is about the automatic creation of hyperlapse videos: We present a method for converting first-person videos, for example, captured with a helmet camera during activities such as rock climbing or bicycling, into hyperlapse videos: time-lapse videos with a smoothly moving camera. So we are now only speaking about first-person videos. As...