https://medium.com/@davincikr1123/release-of-davinci-projects-mainnet-401f12e239baAbstract
Davinci developed DAI (Davinci AI), which is DAPOS type of Chain Generation AI Solution. DAI learns information in MainNet and automatically generates new Sidechain based on the educated characteristics of these nodes. During this performance, it improves the efficiency of the chain operation.
In other words, it means that Sidechain is created automatically by specific DApp similar to ‘Cryptokitties’ and ‘Eos Knights.’ By separating hard and light users, the user’s transaction, contract, and other information are driven in the creation of Sidechain, which reduces the volume of MainNet.
Through DAI’s constant analysis and learning of the transactions, autonomous DPOS based Chain Framework is developed that automatically generates Sidechain to area where the expansion is needed in existing Main Chain (Root Chain or Base Layer). It is possible to obtain improved performance in terms of transaction processing speed, user information security, and blockchain system operation stability via DAI.
Introduction
Davinci Artificial Intenlligence(DAI) 1
Davinci Artificial Intelligence(DAI)’s node analysis is based on following criteria. The volume of MainNet the node belongs and its increasing speed in the number of nodes, the data of average time it takes to select a block generator or a witness through DAPOS method, the data of average time it takes for a block validation node to perform check and to broadcast the result to the entire network, selection of a specific node according to the degree of power and the range of fluctuation within the chain of each node, the unfair trade transaction inspection, and the log activity data of double payment attempts within the chain are collected to be analyzed. DAI models case to reflect the data analyzed through this process for the chain to be generated.
DAI analyses countless nodes as described above and models them. Through the monitoring, when the transaction-per-second (TPS) of the MainNet drops, DAI automatically begins to create a new Sidechain with nodes previously modeled by types. Additionally, the number of transaction occurrence, contract conclusion speed (= TPS), the contract transmission speed between the MainNet and Sidechains, and its stability are monitored
Grouping nodes by type through DAI, a profit model suitable for the characteristics of the node group can be predicted and established when Sidechain is formed and operated. Prediction of business models is provided as visualized data and can be specifically used. Additionally, apart from the generation of Sidechain, it is possible for DAI to create Sidechain with customization of the DApp operator, thereby producing a business model that incorporates operator condition.
In conclusion, development of DAI, automatic Sidechain generation solution based on MainNet, makes blockchain and service operation more efficient and productive than before.
Davinci Artificial Intenlligence(DAI) 2
Store every transaction on MainNet. Data analysis of store transaction log. DApp creates log. Each log is stored and analyzed.
1) Size of MainNet: netSize
2) Time it takes for number of nodes to reach a certain point: time_NodeCnt_to_ainCnt(100)
3) Time it takes for a block generator to be selected: time_BlockGeneratorSelected
4) Time it takes for a validation node to check: time_Check_BlockValidation
5) Time it takes to broadcast the result to the entire network: time_broadcastingEntireNet
6) The degree of power each node has in its chain: degree_PowerInnerNet
7) Detecting data of unfair trades or double payment attempts: Cnt_Irreg_type1/2
Categorize nodes and case modelling by transaction volume or number of coin holding.
Divide and analyze the log file and count each node’s transaction volume with DAI. The node transaction log is separately analyzed to extract the number of tokens by decrypting the encrypted value of the common key value among the encrypted data recorded in the transaction. Define the rank and the type of the node based on transaction volume and number of tokens.
Node Selection
It finds individual node connection based on the contract address of the nodes scattered around the entire MainNet. Then discovers filtered nodes and lists up the nodes based on the characteristics of each node.
Transaction volume value and token retention are different per nodes. DAI determines the rank of each node these two factors and displays it in alphabet forms. Additional evaluation factors other than the above two factors specify the disposition of each node.
Automatic generation of Sidechains with selected nodes.
Logistic Regression is repeatedly performed using selected nodes. Through this exercises, P1, P2…Pn-1 value nodes are selected and classified as a nodes capable of generating a new Sidechain.
During this process, DAI is formed with 16 hidden layers, which consists of 2 sets of 8 layers. Each layer learns through abstraction. DAI automatically binds the specified inter-node contracts to complete Sidechain.
Logistic Regression
Statistical technique used to predict the probability of an event using a linear combinations of independent variables. Similar with the objectivity of a typical regression analysis, the purpose of Logistic Regression is to demonstrate the relationship between dependent and independent variables as a specific function, as the objectives of a typical regression, and to use them in future predictive models.
Node data is validated by the node validator. The data is broadcasted to DApp of the existing MainNet or contributed as a reference value to generate other node data.
Generation of Sidechain & Application of business model
Node groups are classified with various criteria for generating Sidechain, and each node group’s characteristic can be subdivided into a major and minor features based on information of individual node learnt with CNN methods. Major characteristic is extracted again and goes through automatic generation of Sidechain until it converges into one unique characteristic. The converged characteristic of node becomes a predictable business model.
Davinci MainNet and DAI
DAI extracts a value of each transaction data (from, to, gas, gasPrice, input data, nonce, transaction-index, etc.) and matches the user information of the corresponding node stored in DApp. In other words, first, it extracts data that can be matched with the information collected through DApp from many key values. Then it performs the operation of decoding and matching separately. Below left is the transaction data image. Image on the right is about the data analysis. Data used for the analysis are collected and matched via DApp among encrypted data. Transaction ID, transaction occurrence time (=sign up date), contract address (from, to), value transferred through contract, type of token, name, and amount of token possessions are the data items that can be specified.
DAI digitalizes the characteristics of analyzed nodes and computes the rank of each node. DAI also comprehensively accounts MainNet size, transaction creation time, validity verification time, its recognition level in the MainNet, amount of retained tokens and transaction volume to entitle each node’s rank and type.
Data analysis of node’s transaction
Calculate node rating and node characterization criteria.
Grouping of node
After grouping the nodes by each rank, DAI automatically generates the Sidechain composed of each group. During the Sidechain creation, DAI automatically proceeds the process with contract address (from, to address) at initial chain creation stage to acquire the contract that are necessary for the completion.
Lighten of MainNet through Sidechain
Transaction level comparison diagram of Sidechain and existing MainNet
This feature of DAI reduces the weight of network when a new Sidechain is created with extracted node group and increases overall network transaction speed. While the transaction is in process, ‘Transaction Endorsement Time’ occurs at the same time. In the process, as the smaller the number of nodes to reach the ‘Endorsement’ phase, it takes less time. This principle applies to the case of DAI.
Such a new chain, which is resulted from one characteristic automatically generated by DAI, allows for the creation and development of new predictable business models as node characteristics are taken into consideration, which can be converted into visualization data
Future of Davinci MainNet and DAI
DAI is able to realize the business model according to the intention of the operator by recognizing and classifying the characteristics of the node through machine learning and creating a new Sidechain.
The graph above shows what results can be obtained by increasing the number of selected nodes with the purpose of time. MIDI (Marketing Intention to Drive Income) node number value converges overtime into a single value, which is a group of nodes with one characteristic.
AOA and Davinci Foundation choose to adapt learning and creating a new MainNet through AI. Davinci Foundation has set its aim of achieving the improvement through DAPOS method and carried out this development. Several experiments have demonstrated that new business models can be developed by generating specialized Sidechains, rather than simply aiming at improving speed of automatic generation of Sidechain through artificial intelligence.
Fig 20. Result graph:
1) Total average TPS only operated by MainNet and TPS graph after the generation of Sidechain
2) A graph comparing before (left) and after (right) DAI introduction. Time comparison graph for selecting the same 400 block generators.
By making this even more advanced, Davinci Foundation is going to produce an automatic Sidechain generation machine that has degree of completion, which will contribute to the stability and to development of the blockchain ecosystem.
Davinci Official Website:
https://www.davinci.vision/ Davinci Official Telegram Channel:
https://t.me/davincifoundationDavinci Official Telegram Room:
https://t.me/Davinci_EN
