Guimarret

Foundation Hello again. Continuing the neural networks trail, today we are gonna take a look at the CNN. I expected to cover more on neural networks, but it got too long, so the rest is gonna be covered in future posts. I didn’t explicitly say it, but the last post was mostly the technical and mathematical visualization, focused on building intuition on the matter. The actual application in real life gets more difficult because we also have to be aware of optimizing the training setup, otherwise it’s just unusable. If you put yourself back in the late 70s, which is when they were researching and testing the techniques from the last post, the hardware and storage were just too expensive, and for context it is really easy to end up training models that get massively big. ...

So, this is an extensive and sometimes boring matter but also essential. I’m going to start with latches and flip-flops and hopefully end in the RAM/DRAM and the newer memory systems. Latch/Flip-flops For a fresh start and just for contextualization, latches and flip-flops are bi-stable memory elements. That means they can hold binary information depending on the state. The difference between the two is that the latch is level-sensitive, it stays transparent and keeps the input and output in check the whole time the enable is high, while the flip-flop is edge-triggered, it only captures the value on the transition (the edge) of the clock. ...

I just had a uni class about natural language processing and it lit up the need to write about it to reinforce my learning in the topic Me 2 weeks in the future here, I didn’t expect this post to become this big and didn’t cover everything I wanted. In the future I’ll also cover CNN, RNN, Transformer, and so on. After finishing this post, I thought maybe it’s not for everyone because it became a bit too technical with heavy math (I gave my best in simplifying, but I still have a lot to learn, this is the most I’ve written in one shot in my life until now). So for those who came here just to peek, here are some recommendations: ...

So, after a long time, I’m back to show you what I’ve been studying these past few days. This time, I read Crafting Interpreters and the Tiger book (Modern Compiler Implementation) to get a better understanding of how compilers and programming languages work. Disclaimer: There’s a lot of gaps and unlinked points but it will get better for other posts. Introduction So far so good, I started with reading one chapter a day of the Crafting Interpreters, which was really noob friendly (as clearly intended), I thought that was good, since I knew close to nothing about the topic. After finishing, I did the same with the Tiger book, which was delightful, even though it was immensely harder and much more theoretical. There is still a LOT that I need to learn about, but at least I have a slight idea of how to research it haha. ...

Using the transistor base in here: transistor. I’m gonna talk about logic gates and some basic possibilities like adders, subtractors, multipliers, and dividers. The main idea in these basic posts is to trace the way to assembly instructions and so on. These logic gates are the lowest level we can reach in hardware processing. Obviously, these instructions alone don’t create anything really meaningful, but the correct combination can create computers and processing machines of any type (except quantum computers, but we’re not gonna talk about them today) ...