How can consciousness be explained using the most current physics model of the universe without metaphysical explanations?
Addressing the point of the rise of consciousness from inanimate matter, we may use the science of entropy and negentropy to justify the rise of order from relative chaos, with a steady input of solar energy, given a very long period of time. In other words, evolution as such is not actually a violation of the second law of thermodynamics, in contrast to the tenets of creationists.More to the point of the nature of consciousness, i.e. the spontaneous and creative nature of it as a physical sort of phenomenon — we can make use of the quantum uncertainty and the entanglement of subatomic particles in quantum mechanics in order to provide a conceptual basis for the type of unpredictable and spontaneous behavior we observe not only in humans but in animals as well and even in inanimate molecules and systems.Natural selection would certainly favor an organized system that is responsive to and aware of its environment. In addition, the type of “group consciousness” we observe in the human family as well as many animals and even plants is likewise advantageous for the survival of the community.Thus, given enough time, environmental exigency, and added energy, it ought to be possible to come up with the trappings of organized molecular structures, what we call life, and deliberate purposeful activities organized in intelligent ways, which we call consciousness.
Should I write Random Forest code on my own or use existent implementations?
Sure! Why not? If you're interested in both practical and theoretical understanding any algorithm, one of the best ways to learn it is write it yourself. The nice thing about random forests (and just Google "papers on the implementation of random forests", or possibly go to arxiv and search similar terms, or, for something relatively well understood like RF's, you can try Wikipedia) is that they're quite well understood and there are many implementations. So not only can you code your own, after you've done so you can both check your results and look at the way they've been implemented before to better understand them.I've made a special study of trees and forests. One of the reasons I've done so is because during one of my contracts, I wanted to use a random forest for something. And I could only use R and C and not anything else (really, really long story). It was some time ago, and I wanted to test the usefulness of my imputations. But I was stuck, since the randomForest R package doesn't allow for missing data. (I know it imputes now, and various other things.) What I wanted was a random forest algorithm that would allow for missing values via surrogate splits like regular trees.So I wrote one. (OK. Disclosure time. I actually ripped apart the randomForest - mostly to make sure I was doing the right things in the right ways - and, more importantly, the rpart code as much as possible and tweaked rpart as much as I could for minimum time spent but maximal adherence to actually creating a random forest out of rpart trees.) It didn't have to be terribly efficient code, which was good, because I didn't have all the time in the world to write it, and it was a proof of concept for not huge data.And even though I had really studied trees and forest before, my understanding of them really improved.Even if you never use it, because there are other implementations already that are optimized, you'll understand them better. And who knows? That understanding could lead to a better version of the algorithm.