Mico

One of the best thing about blogging my learnings at micmi.co is that I use it as my notes as I continue my learning. Big shout out to @tommcfarlin and @salcode for inspiring me to just blog!

To those who want to are like me and studying machine learning,

Weekend reading for the tech geeks! 🤓 I just finished a massive write-up on how to build an LSTM neural network from the ground up using C++. I've learned a lot!!! Check it out.

scribbles

Studying for my next blog: LSTM Forward Pass & Backpropagation. Writing by hand is just different. Just with this single page I learned new things and corrected a lot of my flaws: - Stitching trick - Broadcasting - How to align the shapes to make the math work Pick up your pen

I wrote a new post on the math and mechanics of LSTMs. I explain how neural networks "remember" and "forget" using a fun RPG analogy to make it easy to grasp. I hope you enjoy reading this as much as I enjoyed writing it!

Learning LSTM (Long Short - Term Memory). Excited to write a blog post about this soon!

Love math, machine learning, and code? I just published "Writing Fraud-Detection using RNN from Scratch" Have a read and let me know your thoughts! nitter.cf/_micmico/status/210850…

I think it depends who will be responsible if something fails. As long as a human is held responsible, then a human should at least be in the loop. Hence, having a human is never a waste.

With OpenAI disabling x20, I am really tempted to get another Claude x20. You guys know of an easy way to change accounts / profile?

You can delegate a lot of things to AI, but never your pursuit of knowledge and understanding, or your drive for excellence.

It breaks my heart to see my father physically weak and fragile because of cancer. - the one who carried me from our car to the bed when I was little. - the one who wakes up first and sleeps last from work when I was studying to give our daily needs. Our family is optimistic of his recovery. But it's just painful during this time. I want to honor my parents more. To obey them more. To be with them some more. To give back to them some more. To love them more.

"The purpose of computing is insight, not numbers." Short history of computers - hardware: Hand calculators: 1/20 ops. per sec Relay machines: 1 op. per sec typically Magnetic drum machines: 15-1,000, depending somewhat on fixed or floating point 701 type: 1,000 ops. per sec. 1990s computer: 10^9 In one day there are 60 x 60 x 24 = 86,400 seconds. A year = 3.15 x 10^7 seconds. 100 years = 3.15 x 10^9 seconds (greater than the average lifetime). This shows that in about 3 seconds, a machine doing 10^9 floating point operations per second (flops) will do more operations than there are seconds in my (and your) lifetime, and almost certainly get them all correct. From Learning to Learn, Chapter 3

- It has rarely proved practical to produce exactly the same product by machines as we produced by hand. - Simulations are cheaper than real labs. - The people at the bottom do not have a larger, global view, but the people at the top do not have the local view of all the details. Both views are needed to make a good decision. Some tidbits from Learning to Learn Chapter 2.

"The unexamined life is not worth living" - Socrates The main difference between those who go far and those who do not is some people have a vision and the others do not and therefore can only react to the current events as they happen. Learning to Learn, Chapter 1

Information is abundant. Wisdom is rare. Pursue wisdom.

But how can they be certain of this accuracy?