Welcome to my blog. My name is Andrea Muttoni and this is a personal space dedicated to all sorts of topics: from machine learning to music composition, and everything in between.

Posts

• 

  Implementing a Deep Neural Network from Scratch - Part 1

  In this series we build a Neural Network from the ground up, covering basic concepts along the way. In Part 1 we will cover some fundamentals, create a simple baseline algorithm and build our first linear model.

  Dec 28, 2020

• 

  Implementing a Deep Neural Network from Scratch - Part 2

  In this series we build a Neural Network from the ground up, covering basic concepts along the way. In Part 2 we add a better loss function and transform our Linear Classifier into a real Deep Neural Network.

  Jan 1, 2021

• 

  Using GraphQL to Build My Own Cloudflare Analytics Page

  CloudFlare has a nifty GraphQL API that lets you fetch analytics data older than the default 30 day window it shows in the dashboard. In this post I show you how to set it up, and fetch total requests and unique views.

  Apr 15, 2021

• 

  How to Ensure Consistent LightGBM Predictions in Production

  How hard can it be - just use pickle! Here's what I learned deploying an LightGBM model to production in order to ensure consistent predictions.

  Jan 22, 2021

• 

  The Tale of the Stochastic Markdown Bug

  Headings or subheadings randomly not rendering in markdown? In this post I recount an insidious markdown bug that took me much way too long to track down.

  Jan 19, 2021

• 

  AWS for Machine Learning Notes Part 1 - Streaming Real-Time Data with Kinesis

  A summary of how to deal with real-time data in AWS. We will cover the Amazon Kinesis service offering including Data Streams, Firehose, Video Streams and Data Analytics.

  Jan 18, 2021

• 

  Implementing a Learning Rate Finder from Scratch

  Choosing the right learning rate is important when training Deep Learning models. Let's implement a Learning Rate Finder from scratch, taking inspiration from Leslie Smith's LR range test--a nifty technique to find optimal learning rates.

  Jan 8, 2021

• 

  Understanding DataBlocks and DataLoaders in fast.ai

  What are they and how do they work together? In this post we will go over what DataBlocks and DataLoader(s) are at a high level, and explain how they work together in the context of building a Machine Learning model.

  Dec 26, 2020