Skip to the content.

Geo spatial queries with elasticsearch

Other than PostGIS, Elasticsearch has the similar functionality to do geo spatial queries. Here I am not going to compare between these two technologies, but to my limited knowledge, elasticsearch could be used for geo spatial search, however it should not be used to store persistent data as opposed to PostgreSQL and PostGIS. I will focus instead to how to index geo spatial documents / data and query it.

Read More

Tags:

How to get elasticsearch up in docker-compose with colima

Elasticsearch is a great tool to enable search within, not only text data, but also multifield objects. Colima is an alternative to docker, since docker desktop has changed their license to be more commercial focus I have to search for alternative and I found that Colima is one stable alternative for mac. In this guide I will introduce a quick start how to get elasticsearch up in local development environment with docker-compose and colima.

Read More

Tags:

Speechbrain CTC parameters

Speechbrain is a framework to experiments with neural network in speech. It has components and structure that is usually used for speech tasks, and nicely integrated with pytorch. They have lots of recipes / quickstart configurations for certain speech tasks, but their documentation is not there yet. In this article I’ll explain my exploration for speechbrain parameters, especially for automatic speech recognition (ASR) with connectiionist temporal classification (CTC) loss.

Read More

Tags:

Introduction to BLEU in Python

BLEU is a metric to quantify effectiveness of an Machine Translation (MT). It stands for BiLingual Evaluation Understudy $^{[1]}$. In general it solves the problem of different human translation references by different annotators when comparing to machine generated translation. Let’s start from using BLEU in NLTK translation package $^{[2]}$

Read More

Tags:

Productivity Tools

This is my personal choice for productivity tools, you may find these tools usefull but you may also find differently. I use ubuntu desktop and android phone, so my personal choices will based on these platforms.

Read More

Tags:

Deep Learning - Week 3 Lecture Notes

Basic Architecture of Neural Network

graph LR subgraph input layer X1[x1] X2[x2] X3[x3] end subgraph hidden layer H1((h1)) H2((h2)) H3((h3)) end subgraph output layer YHAT((yhat)) end X1 --> H1 X1 --> H2 X1 --> H3 X2 --> H1 X2 --> H2 X2 --> H3 X3 --> H1 X3 --> H2 X3 --> H3 H1 --> YHAT((yhat)) H2 --> YHAT H3 --> YHAT
Read More

Tags:

TLDL Radical Candor Ep5 - Career Conversation

This is a too long didn’t listen (TLDL) version of radical candor ep.5 summary. As a reminder for myself in the future and a person who prefer reading than listening because I quickly forgot. NB: when I use “people” here it usually means software engineers.

Read More

Tags: