Plurrrr

Transformers architectures are the hottest thing in supervised and unsupervised learning, achieving SOTA results on natural language processing, vision, audio and multimodal tasks. Their key capability is to capture which elements in a long sequence are worthy of attention, resulting in great summarisation and generative skills. Can we transfer any of these skills to reinforcement learning? The answer is yes (with some caveats). I will cover how it’s possible to refactor reinforcement learning as a sequence problem and reflect on potential and limitations of this approach.

If you spend a lot of time typing plain text, writing programs or HTML, you can save much of that time by using a good editor and using it effectively. This paper will present guidelines and hints for doing your work more quickly and with fewer mistakes.

The open source text editor Vim (Vi IMproved) will be used here to present the ideas about effective editing, but they apply to other editors just as well. Choosing the right editor is actually the first step towards effective editing. The discussion about which editor is the best for you would take too much room and is avoided. If you don't know which editor to use or are dissatisfied with what you are currently using, give Vim a try; you won't be disappointed.

Last year I lead an engineering team for 5 months. During this time we practiced SCRUM, which enabled us to predictably accomplish what we committed to. Predictable delivery is essential for mid and long term projects, where missing commitments could put deadlines at risk. This post outlines what we achieved, how we operated, and why it matters.