Recently I’ve found myself wanting to dig a bit deeper into PyTorch to really understand how it works and hopefully figure out how to expand on my skillset to try new things. To aid this, I’ve been following the Deep Learning with PyTorch book by Eli Stevens, Luca Antiga and Thomas Viehmann. The book itself has been great and highly useful. I recommend it to anyone. ...
Recently with my consulting, I’ve been delving into Google Cloud, and the wide array of offering they have. Getting started with Deep Learning on the cloud is a cost-efficient method way to utilise GPUs. But there are a few gotchas with cloud providers. This post I wrote here, goes through how to set up a Google Compute instance for deep learning toolkits such as Tensorflow, PyTorch, etc. Read the full post here ...
There is a lot happening in New Zealand’s economy currently. New Zealand is officially in a recession with new GDP figures that came out yesterday, a monetary policy update is due next week and an election is taking place mid next month. Thursday’s numbers from Statistics New Zealand represented a -12.2% reduction in GDP for the June 2020 quarter - the largest quarterly fall in GDP on record. Which, unlike the previous quarter’s figures only captured a week or two of New Zealand’s nationwide 4-week lock down, June 2020 figures captures the lion’s share of the lock down. ...
In email marketing, there are a lot of metrics that can potentially be tracked - it’s hard to tell the wood from the trees. When it comes to tracking it’s important to know what metrics to track, what these metrics mean, and most importantly how the metrics can be influenced. In this post, I’m going to break down the ‘vanity’ metrics in email marketing, more specifically: Open Rate Click through rate. Open and Click-through rates are often referred to as vanity metrics as they’re metrics that measure success of a campaign, without actually measuring the business objective. And, let’s be honest they’re satisfying. ...
Recently, I’ve been involved in experiment design and measurement - specifically AB Testing. This experience has encouraged me to learn more about experimentation because of unique challenges faced with conversion optimisation. This post, goes hand in hand with lead score experimentation. In brief, the unique challenges posed were: Dealing with largely a non-technical audience Low sample sizes Costly human interventions as a variant. Because of this, I chose Bayesian AB Testing as a measurement framework. In brief, without bashing Frequentist methods, some benefits of Bayesian AB testing are: ...