Posts

Showing posts from September, 2021

Getting the MIPI DBI screen for the Sipeed M1S dock to work on Linux

Image
Background What started out as an activity to clear out unused stuff from my cupboard led on to a small project to get Linux working with the MIPI DBI screen for the Sipeed M1S dock. I first got hold of the Sipeed M1S dock back in December 2022. What tempted me was the rich set of features it supports, all in a really small package -- NPU, WiFi/BT/Zigbee, 3 RISCV cores courtesy of the BL808 SoC from Bouffalo Labs. It was also touted as being able to run Linux owing to one of its RISCV cores having an MMU. Beyond playing around with the SDK and a few examples, I didn't venture further due to poor documentation. Quickly, it went back to a box I've conveniently dedicated for dev boards which were marketed to support all kinds of features but disappointingly lack good documentation and were thus chucked aside until I had more time to mess around. That time finally came, in a period when work has gotten mundane and I needed some mental stimulation. In the process of declutter
Post a Comment
Read more

Basics of Integration using Monte Carlo

Recently, someone asked me about Monte Carlo. So, I thought I should write this post to provide a basic introduction of performing integration using Monte Carlo. "Why integration?", you asked. Well, this is because integration is one of the main operations done in computing the posterior probability distributions used in machine learning and probabilistic filtering (e.g. Bayes filter). For example, consider the typical posterior probability expression in the Bayes filtering context, \(p(x_{t} \mid y_{1:t})\), i.e. the probability of hidden state \(x_{t}\) given the observed sequence of measurements up to the current time point \(y_{1:t}\), $$p(x_{t} \mid y_{1:t}) = \frac{p(y_{t} \mid x_{t}) p(x_{t} \mid y_{1:t-1})}{p(y_{t} \mid y_{1:t-1})}$$ The \(p(x_{t} \mid y_{1:t-1})\) is of particular interest. It can be seen as the prediction of \(x_{t}\) from previous observations \(y_{1:t-1}\) and can be expressed as $$p(x_{t} \mid y_{1:t-1}) = \int p(x_{t} \mid x_{t-1}) p(x_{t-1}
Post a Comment
Read more