Blockchain Technology: Neural Consensus Algorithm- New Kid on the Blockchain?

Blockchain is heralded as a revolutionary technology that promises to have the same impact as the internet did twenty years ago. Blockchain’s initial application in fintech, Bitcoin, has given the technology worldwide publicity and created an initial mass following- closer to mass global hysteria.

Next, it was the turn of Ethereum to generate a wider application pool for blockchain technology, driving it far beyond the use case in finance. Ethereum is a open source blockchain platform that developers can use to build distributed applications, launch their own cryptocurrency flavor, and facilitate the exchange of money, creative content, property, shares, or any asset of value. Ethereum introduces the concept of autonomous decentralized organizations (DAO)where organizations can run independently, devoid of humans.

However, despite all the exciting technological advancements, blockchain transactions are still relatively slow. How can it scale to deal with the insane amount of transactions per second (TPS) with wider adoption? Currently, bitcoin handles 7, Ethereum 15, and Ripple 1,500 TPS respectively. Compare this to Visa, which handles 24,000 transactions per second. The blockchain consensus mechanism is largely responsible for slowing things down, whereby each block has to be verified before being granted permanency. Examples of current consensus algorithms-

• Proof of work (PoW): Has security and scalability but lacks efficiency due to its inherently high computational overhead.
• Proof of stake (PoS): Is able to meet the demand of efficiency but lacks scalability and security, for example, by being at risk of the possibility of “nothing at-stake” attacks.
• Delegated proof of stake: Lacks security by utilizing only a small number of decision-making delegates.
• Practical byzantine fault tolerance (PBFT): Lacks scalability, as network overhead increases rapidly with each additional node.
• Hashgraph: Is scalable and efficient but has relatively large confirmation delays due to its structural and procedural constraints, such as the
  connectivity and partition of network nodes.
• Algorand: Is scalable, secure, and efficient, but has strict requirements on the overall connectivity of the network, so the time it takes to reach consensus becomes unpredictable in the case of a network partition.

Neural Consensus is a new kid on the block from SeeleTech. It uses the principle of repeated incomplete random sampling to enable all the nodes in a network to arrive at a consensus and transforms voting by changing it from the current model (discrete) to continuous voting. The time it takes for the network to reach consensus decreases linearly with the number of nodes, making Neural Consensus a scalable algorithm. In testing, a 100,000-node network environment, transactions per second (TPS) reached 100,000 and the transaction confirmation time decreased to just a few seconds. Neural Consensus uses ε- Differential Agreements (EDA) to converge a network upon a singular correct value. EDA is very robust in terms of the overall connectivity of the network and can function even if the network connection is less than 50%. It is also able to handle up to 40+% malicious or failed nodes. Refer to the whitepaper at https://seele.pro/ for more information.

Consensus algorithms are a critical component of blockchain technology and their maturity will impact the future of blockchain. Stay tuned.