Category: General Forum Communication Computers Profession Social Aspects

Author

Message Text For: Human Brain and Neural Network Behavior: A Comparison Navigation: Select Threads Forums

Discussion

Date Posted: Tuesday November 11, 2003 01:24:21 PM On the matter of memory, there is no comparison. Neural networks are potentially faster and more accurate than humans. By John Peter Jesan and Donald M. Lauro Reply Top Bottom Next Previous

JeffStout

Date Posted: Tuesday November 11, 2003 02:15:22 PM Starting the article by stating an old urban legend as fact ("Many studies suggest that humans may use less than 10 percent of their brains' potential power") unfortunately casts a shadow over everything that follows. One of the more exhaustive debunkings of this legend is here: http://www.snopes.com/science/stats/10percnt.htm (A favourite excerpt: "Have you ever heard a doctor say, '...But luckily when that bullet entered his skull, it only damaged the 90 percent of his brain he didn't use'? Of course not.") Reply Top Bottom Next Previous

chris@gerrard.net

Date Posted: Tuesday November 11, 2003 04:44:27 PM I was disappointed in this article. The authors may know a fair bit about neural networks, but the "Human Behavior" section is very weak in terms of credible descriptions of the human cognitive and memory mechanism. I'm not an expert, but have an educated layperson's interest in these areas and found their material sloppy and unconvincing. The thesis that there are degrees of completeness in learning, particularly that completely learned things are never forgotten, while incompletely learned things may/will be is not well supported. Many people forget "1+1=2" due to a variety of causes. As for neural networks being "faster" than human memory, there's no news there. Human mental processes are very slow compared to electronics. More accurate? I hope so, given the malleable nature of human memory. So what's their point? Reply Top Bottom Next Previous

Discussion

Date Posted: Wednesday November 12, 2003 11:13:11 AM Prolonging human brain and neural networks comparison. From the neural networks' point of view [the] human brain can be understood as a set of neural networks, which are not independent. On the contrary they have something in common. Human brain when learning seems to develop the particular skill and at the same time a more generic skill that refers to a class of similar skills. In fact, we can think of learning as building skills of previously learnt skills (both specific and generic). This explains why learning could be so fast for certain individuals at least at certain areas. In case of neural networks there is no transfer of knowledge and skill and every time we wish to teach a neural network a skill we have to start from scratch. best regards Maciek Mlynarski Reply Top Bottom Next Previous

jjesan

Date Posted: Wednesday November 12, 2003 11:27:51 PM Thanks for your feedback. Ofcourse, as you mentioned, we did not go deep on the subject. We wanted to present the information in such a way that a non-technical person can understand easily. As you aware, human-learning is not a simple process. Humans start to learn and build knowledge by recognizing images, classifying new objects, memorizing information, thinking and making decisions, solving problems, collecting data and extracting meaningful information from the collected data and so on. The Human mind is not a simple structure and the procedures applied on the mind structure are not simple. There are still lots of mysteries. But some of the behaviors of human brain are analogous to computers and its algorithmic procedures. The best example for this is Artificial Neural networks. For most of the cognitive mechanism, we have similar computational procedure. If you take classification, we have several classification algorithms. But, still there are some challenges for the cognitive science. Some of them are human's emotion, consciousness, physical environment and dynamic status of human mind. Anyway, our point is, there will be a saturation point in learning certain things. If we learn something and do practice on it, we will not forget it easily. It is stored in our long term memory and when we practice again and again, we become an expert of the area/concept and the information about the area/concept has been burnt into our ROM (our brain). In practical, it can't be volatile. Ofcourse there are possibilities to loose some of that information due to age or illness, but it is acceptable for comparision, because it is analogous to lifetime of the Neural Networks. If the neural network is a hardware, there is a possibility for the hardware to go bad. If it is a software, there is a possibility for a need of more training to meet increasing requirements. Hope, I answered your questions. If not, please let me know. Thanks John Peter  Message edited by: jjesan on Thursday November 13, 2003 12:26:00 PM Reply Top Bottom Next Previous

CHSUAPUKAO

Date Posted: Sunday November 16, 2003 02:42:21 AM Is there an analogy to learning in social networks? Isn't it dangerous and akin to braiwashing? I prefer individual humans with their ambiguities, forgetfulness, re-learning, slowness, and creativity! Steve Torok ------------------------- negotiate123 Reply Top Bottom Next Previous

smendick

Date Posted: Friday January 30, 2004 04:40:46 PM I am no expert in memory or Neural Networks, but I do have a background in Cognitive Science. The way I see it, human memory systems and neural nets have some similarities, but are way different in ways which we can't even begin to understand because of our preconceived definition of cognition. For example, in looking at how people remember things from the past, we'll immediately see that they use artifacts from the world as an aid to the memory that is inside their brains. We will often write things down on paper, save files online, create picture albums, keep old friends around or even listen to music that evokes memory from whenever. In my view memory is not just a process within the brain, it is a complicated construct that owes a significant amount of credit to the way we organize our lives around culture. To say that neural networks are good at many of the kinds of memory that humans are good at is naive at best. I will grant that neural nets are good at remembering exact pictures, learning long sequences of numbers, or recognizing certain patterns quickly, but there is no way we can begin to even compare the two memory systems ( human vs nn ) because of their significantly different functions and uses. We really need to redefine ¿memory¿ to have a real conversation. Reply Top Bottom Next Previous

kha0z

Date Posted: Wednesday April 21, 2004 09:15:05 AM << Anyway, our point is, there will be a saturation point in learning certain things. If we learn something and do practice on it, we will not forget it easily. It is stored in our long term memory and when we practice again and again, we become an expert of the area/concept and the information about the area/concept has been burnt into our ROM (our brain). In practical, it can't be volatile. Ofcourse there are possibilities to loose some of that information due to age or illness, but it is acceptable for comparision, because it is analogous to lifetime of the Neural Networks. If the neural network is a hardware, there is a possibility for the hardware to go bad. If it is a software, there is a possibility for a need of more training to meet increasing requirements. >> I am not convinced this is so. It is arguable the process of learning and practice develops a datapath that is used for fast and accurate information retreival. Perhaps, an index of some sort. Or even the re-indexing of an existing index. This would help explain the process of losing memory. Lost memory is not always realated to to age or illness (analogous to hardware failure), but the result of a bad index due to the execution of a corupt algorithm or just a very large index where skills that are not practiced get moved to the bottom of the index and become increasingly difficult to retreive. Additionally, practicing a skill can be thought of a caching mechanism where that skill/knowledge/thought is kept fresh and ready for access in liquid memory (the cache), where other not so frequent skills are left in the crystal memory (the disk). Just my $0.02. ------------------------- -- Alonso "kha0z" Robles  Message edited by: kha0z on Wednesday April 21, 2004 09:16:43 AM Reply Top Bottom Next Previous

Navigation: Select Threads Forums

[ACM]   [Ubiquity]   [ACM Privacy Policy]