Community Reviews

Rating(3.4 / 5.0, 16 votes)
5 stars
1(6%)
4 stars
4(25%)
3 stars
11(69%)
2 stars
0(0%)
1 stars
0(0%)
16 reviews
April 16,2025
... Show More
Her ne kadar kitabın arkasında "popüler bilim" yazsa da kesinlikle bir popüler bilim kitabı değil. Daha ziyade John Holland kendi akademik makalelerini bir editöre vermiş ve bunları sadeleştir ve kitap haline getir demiş, editör ise normal cümleleri sadeleştirmiş ama konseptten bihaber olduğu için karmaşık sistemler ile alakalı noktaları olduğu gibi bırakmış.

John Holland'a birçok kitapta çokça referans verildiği için bu kitaba başlamıştım ancak konular iyi bir fizik eğitimi almış bana dahi ağır geldi. Oyun teorisi üzerinden tavlada kazanma stratejileri ve neural network kurulumlarını anlattığı bölümleri bitirdim. Bu süreçte bir complex system algoritması nasıl oluşturulur. Hangi düşünce adımlarıyla inşa edilebilir, matematik dili ağır bir şekilde, açıklıyor ancak kitabın okuyucuya yazılmadığı ve Mr. Holland kendisi çalışmalarını yaparken bize sadece onu izleyerek bu işlemleri takip etme olanağı verdiğini hissetmeye başlıyoruz.

Kitabın emergence ile alakalı noktasına geldiğimde artık çok yorulduğum için sayfalarca anlatılan constrained generating procedures (cgp)'ın ne olduğunu dahi bilmiyordum ve 167. sayfada bırakma kararı aldım.
April 16,2025
... Show More
Critical learning for anyone involved in modeling dynamic/evolutionary systems
April 16,2025
... Show More
John Holland is pretty much the dude that invented the genetic algorithm, and has done a bunch of groundbreaking AI work using bottom-up learning techniques. This book discusses a lot of these things, and how a small number of rules can create emergent higher-level behavior. In particular, I found his discussion of his 'classifier system' quite interesting. This is just a bunch of agents that can read/write strings from/to a common bulletin board. Each agent reads something, does something to it and posts it back, but the system keeps track of history. When an agent can produce a "correct" result in the form of a string, it's payed off, and every agent that contributed to building the final string is also payed -- strengthening these agents. When a new type of final result is needed, there are already strong agents that are good at making building blocks, so the system adapts much faster than a top-down design. Anyway, if you're interested in genetic algorithms, bottom-up learning, AI, or emergence, you should read this book. However, it is kinda dry in places, obviously written by an academic.
April 16,2025
... Show More
Not a light or easy read, even if you choose to ignore the maths in the boxes, but worth the effort. By sticking to a few of the many possible examples and explaining them in detail, Holland shows how complex behaviours can emerge from the repeated application of a few simple rules and how this behaviour can appear directed when constrained. I'm sure field has moved on since this was written, and perhaps some of the ideas in here are out of date or have led to dead ends, but that he doesn't claim to have all the answers is something that makes the book a refreshing change from the gee-whiz popular science books that I seem to pick up more often. I don't usually need to make use of a pencil and paper when reading these days, but didn't resent having to do so in this case.
Leave a Review
You must be logged in to rate and post a review. Register an account to get started.