Machine Learning

I wrote a blog post a while back about using a Clojure machine learning library called Cortex to do the Kaggle Cats and Dogs classification challenge. I wanted to revisit it for a few reasons. The first one is that the Cortex library has progressed and improved considerably over the last year. It’s still not at version 1.0, but it my eyes, it’s really starting to shine. The second reason is that they recently published an example of using the RESNET50 model, (I’ll explain later on), to do fine-tuning or transfer learning. The third reason, is that there is a great new plugin for leiningen the supports using Jupyter notebooks with Clojure projects. These notebooks are a great way of doing walkthroughs and tutorials. ...

Update: Cortex has moved along since I first wrote this blog post, so if you are looking to run the examples, please go and clone the Cortex repo and look for the cats and dogs code in the examples directory. There is an awesome new Clojure-first machine learning library called Cortex that was open sourced recently. I’ve been exploring it lately and wanted to share my discoveries so far in this post. In our exploration, we are going to tackle one of the classic classification problems of the internet. How do you tell the difference between a cat and dog pic? ...

After having spent some time recently looking at top-down AI, I thought I would spend some time looking at bottom’s up AI, machine learning and neural networks. I was pleasantly introduced to @mikea’s core.matrix at Clojure Conj this year and wanted to try making my own neural network using the library. The purpose of this blog is to share my learnings along the way. What is a neural network? A neural network is an approach to machine learning that involves simulating, (in an idealized way), the way our brains work on a biological level. There are three layers to neural network: the input layer, the hidden layers, and the output layer. Each layer consists of neurons that have a value. In each layer, each neuron is connected to the neuron in the next layer by a connection strength. To get data into the neural network, you assign values to the input layer, (values between 0 and 1). These values are then “fed forward” to the hidden layer neurons though an algorithm that relies on the input values and the connection strengths. The values are finally “fed forward” in a similar fashion to the output layer. The “learning” portion of the neural network comes from “training” the network data. The training data consists of a collection of associated input values and target values. The training process at a high level looks like this: ...