Psychon and McCulloch Pitts Neurons
Do you know the term "psychon"? This term, which seems to be related to something in a spiritual and mysterious world like Psychokinesis or Psychedelic, was named by neuroscientist Warren McCulloch. He thinks that there is a simplest, more indivisible basic unit like the atom at the root of the human psychological phenomenon, and he called it "psychon". Although he can usually think of neurons as being the atoms of the nervous system and human psychological functions, he thought that it was difficult to explain the mental functions of human beings with 'neuron' alone.
He studied philosophy and psychology, and later majored in neurology. He chose the path as a basic science researcher instead of a clinical doctor, and he studied mainly the pathogenesis of Parkinson's disease. He began thinking about "psychon" because he felt that there was something lacking in explaining human psychological phenomenon by the action of neuron. The paper he met at that time was Allan Turing's historical paper "On Computable Numbers" published in 1936. And he hoped that Turing Machine could reveal the secret of "psychon".
In Detroit, there was a rumor of a genius who had not been properly attended regular school classes. He accidentally read Russell and Whitehead's "Principia Mathematica" from his neighborhood library in three days, and wrote a letter to Russell about the first part of the book. Russell invited him to Cambridge University at the age of only 12, but the proposal was not accepted. The first time they met was 3 years later when Russell was an exchange professor at the University of Chicago. This boy is Walter Pitts. Russell introduced this boy to Rudolph Carnap, a science philosopher at the University of Chicago. Pitts learned abstract logic from Carnab and learned mathematical biophysics from physicist Nicholas Rasheveky. This has laid the foundation for the enormous achievement that connects him with neurons and mathematics.
The fateful meeting of McCulloch and Pitts began in 1941 when McCulloch left Yale and moved to the University of Chicago. He met Walter Pitts, 25 years younger than himself, this year. Pitts was still homeless, and McCulloch, fascinated by his talent, asked him to come into his house. From this time, the cohabitation of their souls began as a comrade. McCulloch's concept of "psychon" was naturally transmitted to Pitts, and Pitts began to systematize his ideas mathematically and logically.
Their joint research was published in 1943 as a monumental paper entitled "A Logical Calculus of the Ideas Immanent in Nervous Activity" Although there are different opinions according to scholars, this paper is considered to be the beginning of the concept of artificial neural networks. This paper has been quoted extensively, but this paper is very difficult to understand. The key points are summarized as follows.
The structures of these neurons are called "MCP (McCulloch Pitts) Neuron" after them. MCP neuron is also called linear threshold gate since it simply classifies the set of inputs into two different classes. Following diagram by Kiyoshi Kawaguchi illustrates this model very well.
McCulloch's "psychon" had been realized as "MCP Neuron". Although every network of MCP neurons encodes some logical proposition, they had many limitations. The next major development in neural networks was introduced by Frank Rosenblatt in 1958.
References
Cybernetics: Defining Human Beings (in Korean) by Sungook Hong
Walter Pitts on Wikipedia
Warren Sturgis McCulloch on Wikipedia
A Logical Calculus of the Ideas Immanent in Nervous Activity
McCulloch-Pitts Neurons by Michael Marsalli
The McCulloch-Pitts Model of Neuron by Kiyoshi Kawaguch
He studied philosophy and psychology, and later majored in neurology. He chose the path as a basic science researcher instead of a clinical doctor, and he studied mainly the pathogenesis of Parkinson's disease. He began thinking about "psychon" because he felt that there was something lacking in explaining human psychological phenomenon by the action of neuron. The paper he met at that time was Allan Turing's historical paper "On Computable Numbers" published in 1936. And he hoped that Turing Machine could reveal the secret of "psychon".
 |
| Warren McCulloch and Walter Pitts |
In Detroit, there was a rumor of a genius who had not been properly attended regular school classes. He accidentally read Russell and Whitehead's "Principia Mathematica" from his neighborhood library in three days, and wrote a letter to Russell about the first part of the book. Russell invited him to Cambridge University at the age of only 12, but the proposal was not accepted. The first time they met was 3 years later when Russell was an exchange professor at the University of Chicago. This boy is Walter Pitts. Russell introduced this boy to Rudolph Carnap, a science philosopher at the University of Chicago. Pitts learned abstract logic from Carnab and learned mathematical biophysics from physicist Nicholas Rasheveky. This has laid the foundation for the enormous achievement that connects him with neurons and mathematics.
The fateful meeting of McCulloch and Pitts began in 1941 when McCulloch left Yale and moved to the University of Chicago. He met Walter Pitts, 25 years younger than himself, this year. Pitts was still homeless, and McCulloch, fascinated by his talent, asked him to come into his house. From this time, the cohabitation of their souls began as a comrade. McCulloch's concept of "psychon" was naturally transmitted to Pitts, and Pitts began to systematize his ideas mathematically and logically.
Their joint research was published in 1943 as a monumental paper entitled "A Logical Calculus of the Ideas Immanent in Nervous Activity" Although there are different opinions according to scholars, this paper is considered to be the beginning of the concept of artificial neural networks. This paper has been quoted extensively, but this paper is very difficult to understand. The key points are summarized as follows.
- Neurons are in two possible states: firing and not firing, and thus, they define for each neuron i a predicate that is true when the neuron is firing at a given time t: Ni(t)
- They define the solution of a net as a set of logical sentences of the form 'neuron i is firing if and only if' a given logical combination of the firing predicates of input neurons at previous times and some constant sentences including firing predicates of these same neurons at t=0 is true.
- These sentences are a solution for a net if they are all true for it.
- In other words, the sentences describe what the net computes.
The structures of these neurons are called "MCP (McCulloch Pitts) Neuron" after them. MCP neuron is also called linear threshold gate since it simply classifies the set of inputs into two different classes. Following diagram by Kiyoshi Kawaguchi illustrates this model very well.
McCulloch's "psychon" had been realized as "MCP Neuron". Although every network of MCP neurons encodes some logical proposition, they had many limitations. The next major development in neural networks was introduced by Frank Rosenblatt in 1958.
References
Cybernetics: Defining Human Beings (in Korean) by Sungook Hong
Walter Pitts on Wikipedia
Warren Sturgis McCulloch on Wikipedia
A Logical Calculus of the Ideas Immanent in Nervous Activity
McCulloch-Pitts Neurons by Michael Marsalli
The McCulloch-Pitts Model of Neuron by Kiyoshi Kawaguch
Comments
Post a Comment