- formatting
- images
- links
- math
- code
- blockquotes
•
•
•
•
•
-
Don't Drink and Derive: A Simple Proof that 1 = 2
In this blog post, we present a fun little proof that 1 is actually equal to 2. Or is it?
-
CUDA Basics: Decoding DTMF Signals
Discover the fundamentals of CUDA programming through a dual-tone multi-frequency (DTMF) signaling example. Learn how to generate DTMF signals, save them as .wav files, design FIR bandpass filters, and decode dialed keys. This project includes a CUDA/C++ implementation (dtmfCuda.cu) to showcase the power of parallel computing for signal processing.
-
A few Bit-Twiddling Tricks
When working with bit-fields, there are several techniques that can significantly speed up code — beyond the inherent performance benefits that bit-fields provide for many problems. In this post, I will introduce a few bit-twiddling tricks that I occasionally use, which can be useful for a variety of tasks.
-
The Monkey and Coconut Problem
Three sailors and a monkey are stranded on a deserted island. They gather a pile of coconuts to divide the next morning. During the night, each sailor wakes up, divides the pile into three parts, gives the extra coconut to the monkey, hides his share, and restacks the remaining coconuts. In the morning, the remaining coconuts are again divided into three parts, with one coconut left over, which is given to the monkey. The question is: how many coconuts were originally in the pile?
-
Short Notes: Gradient of the Softmax Function for the Cross-Entropy Loss
The softmax function in neural networks ensures outputs sum to one and are within [0,1]. Here's how to compute its gradients when the cross-entropy loss is applied.