Teburin Abubuwan Ciki
Takardu 159 An Yi Bincike
Cikakken bitar wallafe-wallafe daga 2008-2023
49.7% Gano Abubuwan da ba na Al'ada ba
Mafi rinjayen amfani a cikin aikace-aikacen Koyon Injini na blockchain
47.2% Mayar da hankali kan Bitcoin
Babban dandamalin blockchain da aka bincika
46.5% Ayyukan Rarrabawa
Mafi yawan hanyar Koyon Injini
1.1 Gabatarwa
Fasahar Blockchain ta kawo sauyi ga bayyana bayanai da samun su, ta samar da manyan tarin bayanai waɗanda ke ba da dama da ba a taɓa ganin irinta ba don aikace-aikacen koyon injini. Wannan binciken na tsarin taswira yana nazarin takardu 149 na bincike daga 2008-2023, yana ba da cikakken bayyani game da yadda ake amfani da Koyon Injini akan bayanan blockchain a fannoni daban-daban.
1.2 Hanyar Bincike
Binciken ya bi tsarin tsarin taswira kamar yadda Petersen et al. (2015) da Kitchenham & Charters (2007) suka zayyana. Tsarin rarrabuwa ya tsara binciken a cikin muhimman fannoni huɗu: Amfanin, Dandamalin Blockchain, Halayen Bayanai, da Ayyukan Koyon Injini.
2. Muhimman Binciken
2.1 Rarraba Amfanin
Binciken ya nuna cewa gano abubuwan da ba na al'ada ba ya mamaye fagen bincike, yana lissafin kashi 49.7% na duk binciken. Wannan ya haɗa da gano zamba, gano barazanar tsaro, da gano alamu na shubuha a cikin ma'amalar blockchain.
2.2 Nazarin Dandamalin Blockchain
Bitcoin ya kasance dandamalin blockchain da aka fi bincika (47.2%), sannan Ethereum (28.9%) da sauran dandamali. Wannan maida hankali yana nuna girma na Bitcoin da kuma tarihin ma'amala mai yawa.
2.3 Halayen Bayanai
Kashi 31.4% na binciken sun yi amfani da tarin bayanai sama da maki 1,000,000, yana nuna buƙatun ƙarfin girma don aikace-aikacen Koyon Injini na blockchain. Nau'ikan bayanai sun haɗa da zane-zanen ma'amala, jerin lokaci, da sifofin da aka ciro daga bayanan blockchain.
2.4 Tsarin Koyon Injini da Ayyuka
Ayyukan rarrabawa suna jagoranci da kashi 46.5%, tare da tari (22.6%) da koma baya (18.9%) suna bi. Hanyoyin koyo mai zurfi, musamman Cibiyoyin Jijiyoyi na Zane (GNNs), suna nuna ƙara amfani da su don nazarin zane-zanen ma'amalar blockchain.
3. Aiwatar da Fasaha
3.1 Tushen Lissafi
Aikace-aikacen Koyon Injini na blockchain sau da yawa suna amfani da algorithms na koyo na tushen zane. Babban aikin jujjuyawar zane ana iya bayyana shi kamar haka:
$H^{(l+1)} = \sigma(\tilde{D}^{-\frac{1}{2}}\tilde{A}\tilde{D}^{-\frac{1}{2}}H^{(l)}W^{(l)})$
inda $\tilde{A} = A + I$ shine matrix na kusanci tare da haɗin kai, $\tilde{D}$ shine matrix na digiri, $H^{(l)}$ ya ƙunshi sifofin kumburi a Layer $l$, kuma $W^{(l)}$ shine matrix mai nauyin horo.
3.2 Aiwar Lambar
import torch
import torch.nn as nn
import torch.nn.functional as F
class BlockchainGNN(nn.Module):
def __init__(self, input_dim, hidden_dim, output_dim):
super(BlockchainGNN, self).__init__()
self.conv1 = GCNConv(input_dim, hidden_dim)
self.conv2 = GCNConv(hidden_dim, output_dim)
def forward(self, data):
x, edge_index = data.x, data.edge_index
x = F.relu(self.conv1(x, edge_index))
x = F.dropout(x, training=self.training)
x = self.conv2(x, edge_index)
return F.log_softmax(x, dim=1)
# Misalin amfani don gano ma'amalar da ba ta dace ba
model = BlockchainGNN(input_dim=64, hidden_dim=32, output_dim=2)
optimizer = torch.optim.Adam(model.parameters(), lr=0.01)4. Sakamakon Gwaji
Binciken ya nuna bambance-bambance masu mahimmanci a cikin ayyukan Koyon Injini daban-daban. Tsarin gano abubuwan da ba na al'ada ba sun sami matsakaitan maki F1 na 0.78-0.92, yayin da tsarin hasashen farashi ya nuna Kuskuren Kashi na Tsaka-tsaki (MAPE) daga 8.3% zuwa 15.7%. Aikin ya dogara sosai akan ingancin bayanai, ƙera fasali, da zaɓin tsarin tsari.
5. Bincike Mai Zurfi
Takaitaccen Bayani a cikin Jimla Daya:
Wannan binciken na taswira ya fallasa wani fage da ya fi mayar da hankali kan gano abubuwan da ba na al'ada ba a Bitcoin, yana bayyana duka girma na wasu aikace-aikace da kuma manyan gibi a cikin haɗin kai na sarƙaƙƙiya da haɓaka sabon algorithm.
Sarkar Ma'ana:
Binciken ya bi sarƙar dalili mai bayyanawa: bayyana bayanan blockchain → manyan tarin bayanai na jama'a → damar Koyon Injini → yanzu maida hankali kan 'ya'yan itacen da ba su da wahala (gano abubuwan da ba na al'ada ba) → buƙatar haɓaka ƙwararrun hanyoyin sarƙaƙƙiya da sabbin hanyoyin Koyon Injini.
Abubuwan da suka Fito & Matsaloli:
Abubuwan da suka Fito: Cikakken ɗaukar hoto na takardu 159, tsarin hanyar bincike a sarari, gano rinjayen Bitcoin (47.2%) da mayar da hankali kan gano abubuwan da ba na al'ada ba (49.7%).
Matsaloli: Dogaro da yawa akan bayanan Bitcoin, rashin daidaitattun tsare-tsare, ƙaramin bincike na sabbin gine-ginen Koyon Injini kamar masu canzawa don bayanan lokaci, da ƙaramin nazarin sarƙaƙƙiya.
Hankali mai Aiki:
Masu bincike ya kamata su karkata zuwa ga Ethereum da sarƙoƙi masu tasowa, su haɓaka tsare-tsaren Koyon Injini na sarƙaƙƙiya, da bincika sabbin gine-gine. Masu aiki ya kamata su yi amfani da ingantattun tsarin gano abubuwan da ba na al'ada ba yayin tunkarar daidaitawa.
6. Hanyoyin Gaba
Binciken ya gano manyan hanyoyin bincike guda huɗu: sabbin algorithms na koyon injini da aka ƙera musamman don halayen bayanan blockchain, tsare-tsaren daidaitawa don sarrafa bayanai da kimanta samfuri, mafita ga matsalolin ƙarfin girma na blockchain a cikin yanayin Koyon Injini, da nazarin hulɗar sarƙaƙƙiya. Wuraren da ke tasowa sun haɗa da koyon tarayya don bayanan blockchain na sirri da kuma koyon ƙarfafawa don aikace-aikacen kuɗin rarraba.
7. Bayanan Kara Karatu
- Palaiokrassas, G., Bouraga, S., & Tassiulas, L. (2024). Koyon Injini akan Bayanan Blockchain: Nazarin Taswirar Tsari. arXiv:2403.17081
- Petersen, K., Vakkalanka, S., & Kuzniarz, L. (2015). Jagororin gudanar da nazarin taswirar tsari a cikin injiniyan software. Bayanai da Fasahar Software, 64, 1-18.
- Zhu, J. Y., et al. (2017). Fassarar hoto-zuwa-hoto mara biyu ta amfani da cibiyoyin adawa da juna. Gabatarwar taron na'urar gani ta IEEE.
- Kipf, T. N., & Welling, M. (2016). Rarrabuwa mara kulawa tare da cibiyoyin jijiyoyi na zane. arXiv:1609.02907
- Nakamoto, S. (2008). Bitcoin: Tsarin kuɗin lantarki mai amfani da takwarorinsu.