Arken Lussk
Thrills, Chills, and Kills
Metaheuristic Artificial Cognitive Architecture
- Intent: To create a framework for future Aegis Systems projects.
- Image Source: N/A
- Canon Link: N/A
- Restricted Missions: N/A
- Primary Source:
Metaheuristics - Stochastic Optimization
- Neural Networks
PRODUCTION INFORMATION
- Manufacturer: Aegis Systems
- Model: Version 1.1.1
- Affiliation: Aegis Systems, Arken Lussk
- Modularity: Upper-level heuristics, algorithms, etc.
- Production: Semi-Unique
- Material: Binary, other computer coding.
- Metaheuristics - These are problem-independent methodologies. As such, they do not take advantage of any specificity of the problem and therefore can be used as black boxes. They may even accept a temporary deterioration of a solution (simulated annealing), which allows the architecture to explore a greater solution space more thoroughly to achieve efficacy. Metaheuristic capability also provides a set of guidelines or strategies to develop lower level and problem-dependent heuristic algorithms. The cognitive architecture is thusly able to quickly develop procedures to address specific problems after encountering them through trial-and-error methods or even game theory.
- Neural Network / Deep Learning - A neural network is a set of algorithms loosely modeled after the human brain that are designed to recognize patterns and notice anomalies. They can interpret data through sensory means or through manual input in the form of clustering or labeling. Using the input of information of millions of sensors and nodes, the cognitive architecture wields the capacity to create deductions and formulate lower level heuristic algorithms based off of this influx of information.
Strengths:
- [+] Metaheuristics - These are problem-independent methodologies. As such, they do not take advantage of any specificity of the problem and therefore can be used as black boxes. They may even accept a temporary deterioration of a solution (simulated annealing), which allows the architecture to explore a greater solution space more thoroughly to achieve efficacy. Metaheuristic capability also provides a set of guidelines or strategies to develop lower level and problem-dependent heuristic algorithms. The cognitive architecture is thusly able to quickly develop procedures to address specific problems after encountering them through trial-and-error methods or even game theory.
- [+] Neural Network / Deep Learning - A neural network is a set of algorithms loosely modeled after the human brain that are designed to recognize patterns and notice anomalies. They can interpret data through sensory means or through manual input in the form of clustering or labeling. Using the input of information of millions of sensors and nodes, the cognitive architecture wields the capacity to create deductions and formulate lower level heuristic algorithms based off of this constant influx of information.
Weaknesses:
- [-] Scalability / Portability: The sheer processing power required to use the MACA at full capacity is exponential. Nothing short of quantum supercomputers and massive databanks can power a single iteration of this cognitive architecture.
- [-] Physical Nodes: Physical nodes fueling the data into the MACA's neural network are very susceptible to failure. Be it a platform (datapad, droid, etc) that fields Aegis Systems software - some of these platforms are not capable of transmitting sensory or user data to the architecture's processing center at all. Current nodes are few and far between, severely limiting the functionality of the MACA.
- [-] Inaccurate Inputs: Incorrect information transmitted to nodes can easily stump lower level heuristics of the MACA. It's possible for incorrect data to result in inaccurate deductions revolving around specific problems and leading to accidents to occur, such as faulty numbers on how many people are browsing Aegis Systems' holosite or worse.
DESCRIPTION
The Metaheuristic Artificial Cognitive Architecture is by far Aegis Systems most ambitious project to date. Rousing a horde of systems engineers, mathematicians and computer scientists, Aegis Systems has set its sight on the precipice of computer theory in the known galaxy. The end result was a framework capable of being utilized in future projects revolving around Aegis Systems primary mission objective of providing a safe, secure, and stable Holonet for the entirety of the galaxy.
The primary features of the architecture itself are its metaheuristic algorithms. These are upper level algorithms capable of being problem-independent and deducing and reasoning generalized solutions while also providing information to create guidelines and procedures for lower level heuristic algorithms. Effectively, the architecture can efficiency reason through problems and create its own lower level procedures to work through them at the same time. Through the simulated annealing technique, the architecture may even allow situations or solutions to deteriorate to further expands its options in a search area to more thoroughly reason through solutions.
Providing the sensory data that is required to make these deductions is no easy task. That's why Aegis Systems crafted their own neural network capable of deep learning within the architecture. These simple algorithms provide sensory data (machine language) in clustered or labeled formats as raw input, into the architecture's black box base. However, these nodes come in many shapes and sizes. Some instances include Aegis Systems software in their droids or machinery, other examples even include their commercial software suites available to your regular consumer. Many nodes are not capable of transmitting this sensory information and some receive incorrect data, which lead to unreasonable deductions made by the architecture.
Further limitations include the sheer amount of processing power required to even run a single instance of this architecture. Nothing short of a quantum computer and a multitude of high-capacity databanks can fully attain optimal levels of performance from this architecture.
Summarily, this is the first known venture into neural networks and high-level machine learning. Aegis Systems intends on using this cognitive framework as a baseline for future ventures into artificial intelligence and commercially available software suites.
