project

anonymous@undisclosed.example.com (Anonymous) — Wed, 13 Sep 2017 04:33:04 +0000

@@ -2,9 +2,9 @@
  [[people:demitri_maestas|<-- Return to Demitri's page]]
  ----
  
  ==== An Artificial Neuron ====
- Evolution has done so much of the dirty work for us; 3.5 billion years of attempts at creating robust living systems {{ :people:demitri_maestas:project:assets:neuron.jpg?direct&100|}}that work on a scale that begins inside of a membrane in a single cell and stretches out to the stratosphere. An industry-based data storage model in our current machines hold no similarity to the trials and tribulations, which is odd considering the fact that **Mother Nature has 3.5 billion years of design under her belt, where as we have only had a Von Neumann machine for about 80 years.** We need to be thinking about robust storage of information not in memory addresses based on a data structure, but instead, contained within the axons of artificial neurons that make mimic the design structure of the most robust //organic// hard drive we have; The human brain.  
+ Evolution has done so much of the dirty work for us; 3.5 billion years of attempts at creating robust living systems {{ :people:demitri_maestas:project:assets:neuron.jpg?direct&100|}}that work on a scale that begins inside of a membrane in a single cell and stretches out to the stratosphere. The industry-based data storage model in our current machines holds no similarity to the trials and tribulations that these living systems have endured, which is odd considering the fact that **Mother Nature has 3.5 billion years of design under her belt, where as we have only had a Von Neumann machine for about 80 years.** We need to be thinking about robust storage of information not in memory addresses based on a data structure, but instead, contained within the axons of artificial neurons that make mimic the design structure of the most robust //organic// hard drive we have; The human brain.  
  
  ==== Robustness in Machine Learing ====
  We've all seen what machine learning is capable of and the things it can help us do. Just take a look at this awesome machine learning example that Youtuber "SethBling" has designed to play the first level of Super Mario World:{{ youtube>qv6UVOQ0F44?large }}

work_log

anonymous@undisclosed.example.com (Anonymous) — Sat, 30 Sep 2017 06:08:54 +0000

@@ -6,27 +6,28 @@
  ==== 29-Sep-2017 11:36:45PM-0600 ====
  
  Thanks to some old guy who doesn't sleep, I was able to implement an elegant solution for an inversion of the C2D x and y. The following was sent to me by Dr. Ackley:
  <code>
- C2D offset = whatever();
+ C2D offset = someFunction();
  C2D opposite = -offset;
  </code>
  
  So, when a scout hits another neuron, or a Myelin sheath, it will immediately invert direction to report that information back to the neuron.  
  
  **Accomplishments**: 
-   * Heavily worked on the scouting mechanism to help a nucleus determine its relative location compared to the Neural structure (or lack thereof) and determine the best course of action.
+   * Determined that in the spirit of keeping stigmergy a thing, The system will move a little slower than an optimized neuron. 
  
    * Re-formatted my worklog on the wiki
  
  
  **Questions**:
-    * What is the "best" way to work with coordinates? More specifically, what is the best way to work with getting a direction opposite to you in a robust way? Let's say I'm moving in (1,1) each step. How can I determine that I should be going (-1,-1) If I want to switch directions? I can think of some solutions, but they are not very elegant. 
+    * What's a good way to determine how much information I should hold in a single neuron? There is 91 bits in our atom, which means 2^(91) (or something like that) possible combinations of bits in our memory. What if we don't need that much? How do we determine a re-structuring if we find out we need more than we've engineered for? 
  
  **Comments**:
    * I'll make sure to verify that the off-by-one error will fix not affect performance.
+   * You're never too old for trix.
  
- 
+  --- //[[chucho2@unm.edu|Demitri D Maestas]] 2017/09/30 00:08//
  ----
  
  ==== 29-Sep-2017 04:05:40AM-0600 ====

Robust-first Computing Wiki - people:demitri_maestas

project

work_log