LaTeX neural networks

A LaTeX/TikZ/PGF package for drawing directed graphs, such as neural network schematics.

I started a project to create an open-source mid-level Machine Learning textbook, based on some notes from a Caltech course and a Coursera course. The contributions from the community were of poor quality and laden with mistakes, so I eventually terminated the project (having rejected all public submissions). I will work on the book as an occasional project for my own benefit, but it is not a priority any more. I have stopped working on the book.

Neural network implementation of an exclusive-or (XOR) logic gateTo rapidly produce neural network illustrations in the book, I created a LaTeX package to wrap all the TikZ/PGF clutter. The result is a set of LaTeX macros that allow high quality neural-network graphs to be drawn rapidly, and I have since made the package publicly available via my github repository.

Available on CTAN: /graphics/pgf/contrib/neuralnetwork
Download:
LaTeX neural networks package
Latest version (github): battlesnake/neural

Neural network with two hidden layers

Dielectric: Electrical and acoustic spectroscopy via a PC sound card

This program generates a chirp signal, sweeping over a given frequency range, and drives a given audio output with the signal, while simultaneously analysing the signal coming from an audio input.  This allows electrical spectroscopy and acoustic spectroscopy from ~10Hz-100 kHz to be performed with a £40 sound card.  Note that AC97-channels (such as on-board sound) have lowpass filters built in, restricting the upper limit to around 25 kHz.  Tested with a Creative Audigy 2 ZS, using kX v5.1 drivers.

Download here.

test-120k-1k-fullrange