The Tensor Collective

Machine learning is inherently simple. Take a random matrix, apply it to your input and calculate a number that is proportional to how close you are to your target for said input. Now use an automatic differentiation package to calculate how much you need to change the parameters to get a better model. Repeat this enough times and you’re on your way to the vast majority of deep learning algorithms.

Many of the intimidating concepts in ML don’t require degrees in mathematics and statistics to grasp. Better yet, as we embrace the tools of today it helps to understand their technical grounding. With our articles, we bridge the gap between technical aspects and the user experience of these applications. It’s become a bit of a buzzword however and machine learning isn’t fit for a majority of problems. However, the algorithms we use have become extremely efficient and will continue to do so. I think there’s a huge appetite for a platform to onboard people from scientific platforms into the machine learning domain; my primary focus is to provide intuition and insights into the field.

We’ll cover concepts in more extraneous domains too, reinforcement learning and unsupervised learning. Domains rich in application but due to their complexity have perhaps not received as much public attention. ML gives us prospects of a revolutionised society but also presents fears concerning the ongoing rapid race to the bottom for AGI. ML has a lot of “smokescreen” papers, and a contiuous race to state-of-the-art (SOTA) that propagates much of academia. I try to highlight the true lessons we learnt in this regard from the progress of the greatest models we use today.

Motivation: Bridging the Gap between STEM and Robust AI Robust AI algorithms are complex, and requires deep understanding of the underlying principles and limitations.

Clarity in Communication: The Heart of Our Content I seek to break down technical aspects of AI into digestible, easy-to-understand essays with clear narratives.

Visuals: A dedicated approach Great visualizations make an article, I spend a large amount of my time writing dedicated to creating clear visuals that follow a consistent and pleasant markdown.

Article Types: Easy reads and technical pieces As this is a personal blog, I’ll include some personal projects of mine too and the lessons I’ve learnt in doing them – It’s important to be open about the pursuits that don’t work out in ML. I’ll have some easy-reads which maybe don’t cover specific applications but rather insights you mind find in your own pursuits too!

About the Author

I’m finishing my study of Chemistry at the University of Oxford this year. I pivoted to machine learning in recent years with a curiosity founded on how simple machine learning can be to it’s core. We’ve seen lots of applications of ML, perhaps some would argue too many, and increasingly we’ve seen applications across scientific domains. I’m passionate to help other scientists bridge the gaps in their knowledge, like I’ve done myself, to reap the benefits of AI technologies and techniques.

My journey as a Chemist began with an infatuation for the vast complexity of chemical processes, and their wildly unpredictable behaviour. The Briggs-Rauscher displays oscillating changes between vibrant shades of blue and pale yellow, only explainable via a complex chain of interlinked single-step reactions. Algorithms capable of surpassing human-level identification are too, nothing more than a complex orchestration of simple building blocks; simple calculus and matrix products can get you to an accuracy of around 90% for looking at hand-drawn digits.

Join Me

From AlphaFold2 reaching crystallographic accuracy, to DALLE-3 creating realistic imagery in seconds. Deep Blue reigning over the world champion of chess to the superhuman intelligence of the language models in the GPT series. Machine learning will continue to yield powerful algorithms that power the modern world for centuries to come.

Feel free to reach out to me if you want to publish some work on here too. Stay tuned for our upcoming posts!