Benefits of decentralized indicator regression
"decentralization" is translated from the English word decentralization, which is composed of the prefix de -, stem central and suffix ization. Among them, the stem central means "center", the suffix - ization means "... Hua", and the prefix de - has the meanings of leaving, removing, canceling and opposite. Therefore, it is very accurate to translate it into decentralization
what is the meaning of decentralization
vitalik buterin, founder of Ethereum, published the article "the meaning of decentralization" in February 2017, elaborating the meaning of decentralization. He thinks that we should distinguish the centralization and decentralization of computer software from three perspectives: architecture, governance and logic
Architecture centralization refers to the number of nodes that the system can tolerate and continue to run; Governance centralization refers to how many indivials and organizations are needed to control the system; Logic centralization refers to whether the interface and data presented by the system are like a single whole
blockchain is a unified account book of the whole network, so it is logically centralized, which is beyond doubt. From the perspective of architecture, blockchain is based on peer-to-peer network, so it is decentralized. From the perspective of governance, blockchain makes it difficult for a few people to control the whole system through consensus algorithm, so it is decentralized. The decentralization of architecture and governance brings three benefits to blockchain: fault tolerance, anti attack and anti collusion
five differences between blockchain and traditional distributed system
as a new kind of distributed system, blockchain is often mistakenly regarded as a distributed database or log system. In fact, there is a fundamental difference between blockchain and traditional distributed system, namely decentralization. Now let's look at the main differences between blockchain and traditional distributed systems:
(1) consistency algorithm: what blockchain needs to solve is the Byzantine general problem, that is, there are one or more fraulent nodes in the network, which may deliberately violate the protocol or transmit wrong data, Therefore, Byzantine fault-tolerant consensus algorithms, such as BFT, pow, POS, are often used in blockchain; The traditional distributed system only needs to consider the node failure and communication errors, and often uses consistency algorithms such as Paxos and raft, which can not resist the fraulent nodes
(2) central controller: there is no central controller in the blockchain network, and no node can control or coordinate the generation of ledger data. All nodes coordinate through consensus algorithm to generate consistent ledger. However, the traditional publishing system is often controlled by one organization, which uniformly dispatches all nodes to participate in the calculation
(3) rule making: the rule of blockchain is consensus protocol, also known as consensus mechanism, and consensus algorithm is a part of it. Consensus mechanism is generally designed and developed by a person or a team, and the corresponding proceres are developed for the community to use. This seems to be the same as the traditional distributed system, but the change and upgrading of the consensus mechanism of the blockchain requires the community to have a consensus on it. If no consensus can be reached, anyone can implement hard bifurcation and build another community and chain. This is the decentralization process of consensus mechanism.
(1) fault tolerance: once there is a problem in the center, other nodes are easy to collapse. The centralized system is unlikely to be unexpected, because it depends on other nodes, and other nodes can not have problems together
(2) anti attack ability: decentralized system will make the cost of being attacked higher, because it lacks a sensitive central point, and the central point is more likely to be attacked by low cost. The reason is that we should all understand that the attack center may completely collapse, which is why more and more investors want the decentralized technology to become more mature
(3) anti Collusion: it is difficult for the participants in the decentralized system to sacrifice other participants and plot to make their own profits.
The purpose of data centralization and standardization is to eliminate the differences between features, which can make different features have the same scale and make the influence of different features on parameters consistent. In short, when the scale (unit) of the features on different dimensions of the original data is inconsistent, the data needs to be preprocessed by centralization and standardization steps
extended data:
because the original data often have different units of independent variables, it will bring some difficulties to the analysis, and because of the large amount of data, the calculation result may not be ideal e to rounding error. Data centralization and standardization are helpful to eliminate the influence caused by different dimensions and orders of magnitude, and avoid unnecessary errors
in regression analysis, it is usually necessary to centralize and standardize the original data. Through centralization and standardization, the data with mean value of 0 and standard deviation of 1 are obtained
