In NeuroChain, the terms Machine Learning or Artificial Intelligence refer to the high level of abstraction of Bots. They are composed of different categories of algorithms. These algorithms will be used for different applications: consistency algorithm for traceability, Bayesian network algorithm for detecting anomalies in transactions and information emitted by Bots, pricing algorithm for information exchanges and “improved rules-based-system” for complex functional applications. IPFS protocol and cryptographic signatures will be used for the algorithm lineage authentications.

In following, few algorithms in NeuroChain context are described.

Coherence Algorithm:

  • the traceability is based on the continuity of the lifeline of the object or the concept integrated in the Blockchain. The sender of the traceability chain specifies the intrinsic information about the object and its interactions with the Bots in the first transaction. The algorithm verifies the consistency of different transactions, intrinsic metadata and associated certified documents (via IPFS) to validate operations. The algorithm also verifies time coherence and transaction sequences. For example, if an object skips a step in the chain of custody or stagnates in a particular step of the process, the algorithm will detect them as suspicious objects.

Bayesian network algorithm:

  • (Darwiche Adnan, 2009; Gelman, Carlin, Stern, & Donald B Rubin, 2003) is probabilistic graphical model adapted for high dimension and heterogeneous data, and it is based on conditional probability calculation.

In probabilistic graphical model, each node represents a variable (or an observable) and the link between the nodes represents the possible causal relations and correlations between them. Transposing this to NeuroChain, the nodes represent the Bots and the links represent the different transactions between them. The created graph is oriented, and therefore the Bayesian network belongs to the directed acyclic graph (DAG). The threshold for the anomaly detection algorithm will be dynamically fixed according to the network.

Consider the following example: suppose that there are two events A, B which could cause the event C. Suppose also that the event B could impact the event A.

The join probability function is therefore:

 

With Bayesian networks, there are three main inference applications, such as inferring unobserved variables, parameter learning and structure learning. These applications echo basic Bayesian analysis for given data and parameters, with prior probabilities and likelihood. This approach is fully consistent with distributed Bot network.

As an example, this approach is wildly applied in medical research, to calculate the probability of illness depending on the symptoms.

 For the anomaly detection operation other verifications related to the atomic transactions of each Bot will be taken into account.

The idea behind is that the algorithm learns from the transactions between the Bots to calculate the parameters of the model. Then, at each new transaction, new probabilities of interactions are computed. Malicious transactions with an irregular behaviour will be detected (depending on tolerance threshold).

Formal concept analysis:

  • A method to construct a concept hierarchy from a collection of objects and their properties (formal anthology). The formal analysis will be applied to validate methodologies or algorithms by formal logic. This will allow the incoherence and anomaly detection (Wille, 1982).

Semantic analysis

  • The semantic analysis will mainly acts in social applications like social Bots. These algorithms will help the users to interact between them via the Bots. Sentimental analysis, entity recognition or dictionary constitution are examples of procedures that will be used to facilitate communication between human users and Bots, and between Bots.

Rules-based-system (Giarratano & Gary Riley, 1998)

  • A powerful engine to handle conditions and rules. This type of systems is fundamental for artificial intelligence. In NeuroChain, the rules-based-system will be used for smart business applications to store and manipulate knowledge in order to interpret information in pertinent way depending on the application.

Trend detection algorithms:

  • In NeuroChain, Bots are authorized for crypto Value-Barter in the chain. To assess these values, trading algorithms are also available based on the network. New generation of distributed pricers and prevision algorithms will be developed. Trading platforms will, therefore, appear in the system.

Bibliography

Darwiche Adnan. (2009). Modeling and Reasoning with Bayesian Networks. Cambridge University Press, ISBN 978-0521884389.

Gelman, A., Carlin, J. B., Stern, H. S., & Donald B Rubin. (2003). Fundamentals of Bayesian Data Analysis: Ch.5 Hierarchical models. CRC Press. ISBN 978-1-58488-388-3., 120.

Giarratano, J. C., & Gary Riley. (1998). Expert Systems. PWS Publishing Co. Boston, MA.

Wille, R. (1982). Restructuring lattice theory: an approach based on hierarchies of concepts. Rival, I. (ed.) Ordered Sets.445-470.

Invest Smart! Invest NeuroChain!

Click below to join in the conversation

Photo credit: PEXELS 

NeuroChain CES Las Vegas
Comments are closed.