869 private links
This video specifically talked about the real difference in biological neurons and its artificial counterpart. I found the part is on how to represent information and training "artificial biological neuron networks" interesting. Notes (terms: neuron := biological neuron):
- neurons accumulates the incoming charge (stateful), with charge leakage
- three ways to encode information: frequency, timing, and parallelism
- backprop breaks down on discrete signals
- the only known mechanism for neuron weight update is "neurons that fires together wires together". but it's not very useful to developing a way to train it.
- it's likely neurons interpret its input as binary.
- the firing rate limit of neurons is only around 4 ms/spike (very slow)
- directional hearing requires distinguishing 1/2 ms delay of signals. how is it possible given the maximum firing rate? answer: create a group of neurons each detecting different delay and ordering in input signal.
- neuron components:
- loopback neuron can store a bit of signal that can be set and reset. (like an SR latch)
- mpsc (read-write) buffer
- mechanism that repetitively reading refreshes the memory (like DRAM)
- power efficiency. (12W for the brain) calculated result: neocortex only fires every two seconds (Wow!)
A comprehensive video tutorial on the Transformer architecture by Andrej Karpathy.
Find the source of screenshots of videos, etc.
On the techniques required to circumvent YouTube's download rate limit.
A tool to quickly find the CRF setting needed for achieving target VMAF score.
How different bitrate mode works for video encoding and which one should you choose?
In short, CRF should be good enough for most non-streaming use because it doesn't waste bits on easy to compress clips (quality based). Otherwise use a Capped-CRF if a maximum bitrate cap if necessary.
Various bitrate control methods are suitable for different purposes. This article gives a clear description on how to use them in ffmpeg and when to use them.
Inspect mp4 box structure.
A fasterthanlime video on injecting code to a foreign Windows processes.
Unison is a langauge where segments of code are not addressed by names (like function names), but by the hash of their contents represented as AST.
By building a language based on this central idea, the problem of dependency conflict is a non-problem. The author also demonstrated the potential of elastic distributed computing based on the idea.
This project reminds me of From Laptop to Lambda: Outsourcing Everyday Jobs to Thousands of Transient Functional Containers, which is built around a similar idea, but implemented as a computing system instead of a programming language.
<blockquote>How can honeybees communicate the locations of new food sources? Austrian biologist, Karl Von Frisch, devised an experiment to find out! By pairing the direc...</blockquote>
<blockquote>Easily download videos from Youtube, Dailymotion, Vimeo and other websites.</blockquote>