Machine Learning’s Hidden Talent: How Encoders and Decoders Turn Data into Language

Mohit Dubey
3 min readMar 5, 2024

The famed Encoders and decoders well known as the fundamental building blocks in many machine learning tasks, particularly those that involve sequences of data like text, speech, or even video. Here is a crash course on Encoder decoder to someone who is not a math major (also from someone who ain’t one either)

Created by Author using Dall-e-3

Spoiler Alert: If you want a mathematical deep dive, stop now and close the page. I am definitely not the right person.

Before we proceed further I would like to introduce two persona’s

Me (Mohit): Attention to detail, structured, very logic oriented, have zero creative genes.

My brother (Rohit): dreamer, storyteller, bursting with creativity, shaggy hair.

Mohit, the Encoder (Summarizer): Good at Compressing Information and Capturing Context

  • A Summarizing machine. Give me a book, I will summarize the top takeaways from that book. Give me a movie to watch, and I will concisely provide a synopsis of the movie.
  • Highlighting the main character: Like a movie reviwer, the encoder a.k.a. me will highlight the main characters, their motivations, key plot points (including spoilers)
  • Connector: The encoder connects the dots…

--

--

Mohit Dubey
Mohit Dubey

Written by Mohit Dubey

Data | Cloud | ML Platform Architect. Publish articles about technology, investing and occasionally personal idiosyncrasies.