Bytom Adopts the POW Mechanism to Secure Blockchain Assets
The vision of Bytom is to bridge assets of atomic world to that of the byte world. To achieve it, we must guarantee the security and stability of the Bytom system. As such, our developers chose to adopt the proof-of-work system at the first place. But how to make the Bytom mining more energy-efficient while keeps it secure?
POW so far is considered the most fault-tolerant consensus mechanism for a public chain. An attacker can only successfully wage an attack with 51% hashrate. However, other consensus mechanisms like Practical Byzantine Fault Tolerance （PBFT）allows an attack only with one third of the nodes.
It’s nothing new for public chains to adopt the POW system, but they often alter or modify mining algorithms to prevent a 51% attack. If a new digital currency adopts the same algorithm of bitcoin, it’s easy for a bitcoin mining farm to have hashrate of over 50 percent of the new coin when they mine the new coin with a small portion of their hashrate. In other words, a newly born cryptocurrency would be under the control of a big mining site, which is in opposite of decentralization. Therefore, a new alternative coin mostly gives its priority to apply a different algorithm.
Usually, a POW-based public chain would adopt a hashing algorithm, such as SHA-256. The reason POW is mostly adopted for mining is because it’s a one-way encryption algorithm. The intention of POW is to make miners perform a certain amount of computational work where they must alter the input value nonstop to find a right output. The input is actually data of a blockheader. Hashing algorithm keeps miners diligent to secure the whole network.
To secure the bitcoin network, bitcoin miners must keep hashing and they are specifically designed for the scheme SHA256. At Bytom, we design an ASIC-friendly mining algorithm so that miners can mine Bytom directly with AI ASIC chips.
The mining algorithm of Bytom must be able to do two things:
AI ASIC chips-friendly. The hashing process of mining Bytom resembles the hashing algorithm of AI ASIC chips. Therefore, miners can simply mine BTM with AI ASIC chips, which can reduce of the cost of AI chips and thus make the mining process less costly.
Bytom Algorithm must be compatible with what is widely used in AI computing. AI algorithms differ in their computing paradigms, which must be considered in the design of ASIC chips. Thus, the mining algorithm of Bytom must be applicable to most popular machine learning algorithms.
The mentioning of AI tends to remind us of machine learning. As a way to realize AI, machine learning covers areas like artificial neural network, decision tree, perceptron, support vector machines, integrated learning, Adaptive Boosting, dimension reduction and measurement learning, clustering and Bayesian networks. And among them, artificial neural networks represent a classic and widely accepted machine learning method that has a wide scope of application.
This is a typical Mcculloch-Pitts Neuron Model.
As we can see, the artificial neuron receives a set of inputs a1,a2,a3,,,,,an, an offset b and an output t. w1,w2,w3….wn are weight values associated with each input line. Sum is the weighted sum.
Machine learning by its nature is to learn from and make predictions though training data in the data sample.
Multi-layer networks are more needed in real life applications. The training process is actually the process of solving Matrix operations.
The mining algorithm if Bytom involves a lot of matrix. And AI-friendly ASIC chips are specifically improved to solve matrix operations so that miners can mine Bytom with enough AI chips.
A reliable and safe mining algorithm is crucial to the operation of a public chain and one of the most important factors for users to trust it. At Bytom, we work to secure the assets on blockchain and make the mining algorithm stable.