Binciken Tsarin Jijiyoyin Kwamfuta A Rarrabe Akan Blockchain Don Hasashen Kudi na Sirri

Table of Contents

• 1 Gabatarwa
• 2 Bayanan Baya
  • 2.1 Blockchain da Ethereum
  • 2.2 Binciken Tsarin Jijiyoyin Kwamfuta
• 3 Matsalar Hasashen Kudi na Sirri
• 4 Hanyar Aiki
  • 4.1 Morphism na Cibiyar Sadarwa da Ingantaccen Bayesian
  • 4.2 Tsarin Ƙarfafawa na Blockchain
• 5 Sakamakon Gwaji
• 6 Yarjejeniyar Blockchain da Aka Tsara
• 7 Aiwatar da Fasaha
• 8 Aikace-aikacen Gaba
• 9 Bincike na Asali
• 10 Nassoshi

1 Gabatarwa

Ƙirƙirar ingantattun cibiyoyin sadarwar jijiyoyin zurfafa ta ƙunshi maimaita ingantaccen aikin hannu na topology da hyperparameters, yana hana aiwatar da aikin sosai. Littattafan baya-bayan nan suna ba da shawarar algorithms daban-daban na Binciken Tsarin Jijiyoyin Kwamfuta (NAS) waɗanda ke sarrafa wannan aikin ta atomatik. Mun yi amfani da keɓaɓɓen algorithm na NAS tare da morphism na cibiyar sadarwa da ingantaccen Bayesian zuwa hasashen kudi na sirri, inda muka sami sakamako masu kama da mafi kyawun samfuranmu na ƙira ta hannu.

2 Bayanan Baya

2.1 Blockchain da Ethereum

Fasahar Blockchain, wacce aka gabatar da Bitcoin a cikin 2008, tana ba da tsarin littafin rikodin da ba za a iya canzawa ba. Ethereum yana faɗaɗa wannan iyawa tare da kwangiloli masu wayo, yana ba da damar yarjejeniyoyin da ke aiwatar da kansu, waɗanda suka zama tushen cibiyar sadarwar NAS ɗinmu da aka tsara.

2.2 Binciken Tsarin Jijiyoyin Kwamfuta

Algorithms na NAS suna sarrafa ƙirar cibiyar sadarwa ta hanyoyi daban-daban ciki har da ƙarfafa koyo, algorithms na juyin halitta, da ingantaccen Bayesian. Hanyarmu ta haɗa morphism na cibiyar sadarwa tare da ingantaccen Bayesian don ingantaccen binciken tsarin gine-gine.

3 Matsalar Hasashen Kudi na Sirri

Muna mai da hankali kan hasashen farashin kudi na sirri ta amfani da bayanan blockchain na tarihi, littattafan oda, da alamomin ra'ayin jama'a. Bayanan sun haɗa da bayanan Bitcoin da Ethereum na shekaru 2 tare da tazara na mintuna 15 a cikin fasalulluka daban-daban 42.

4 Hanyar Aiki

4.1 Morphism na Cibiyar Sadarwa da Ingantaccen Bayesian

Morphism na Cibiyar Sadarwa yana kiyaye aikin cibiyar sadarwa yayin da ake canza tsarin gine-gine ta hanyar ayyuka kamar ƙara Layer, canje-canjen girman kernel, da shigar da haɗin tsallake. Haɗe tare da ingantaccen Bayesian, wannan yana ba da damar bincika sararin tsarin gine-gine yadda ya kamata.

Aikin sayan don ingantaccen Bayesian ana iya bayyana shi azaman:

$a(\mathbf{x}) = \mu(\mathbf{x}) + \kappa\sigma(\mathbf{x})$

inda $\mu(\mathbf{x})$ shine ma'anar baya, $\sigma(\mathbf{x})$ shine bambancin baya, kuma $\kappa$ yana sarrafa ciniki na bincike da amfani.

4.2 Tsarin Ƙarfafawa na Blockchain

Muna ba da shawarar yarjejeniya na tabbatar da aiki mai amfani inda masu haƙo ma'adinai ke gasa don nemo mafi kyawun tsarin jijiyoyin kwamfuta. Aikin lada shine:

$R = \alpha \cdot \text{daidaito} + \beta \cdot \text{inganci} + \gamma \cdot \text{sabon abu}$

5 Sakamakon Gwaji

Algorithm ɗinmu na NAS ya sami daidaiton hasashe na 94.2% idan aka kwatanta da 93.8% na samfuran da aka ƙera da hannu. Hanyar rarraba ta rage lokacin horarwa da kashi 35% yayin da ake kiyaye aikin da ya dace.

6 Yarjejeniyar Blockchain da Aka Tsara

Muna ƙirƙirar cibiyar sadarwa mara tsari inda nodes ke ba da gudummawar albarkatun kwamfuta ga ayyukan NAS. Kwangiloli masu wayo suna sarrafa rarraba aiki, tabbatar da sakamako, da lada na alama dangane da ingantattun samfuran ingantaccen aiki.

7 Aiwatar da Fasaha

Misalin Code: Aikin Morphism na Cibiyar Sadarwa

class NetworkMorphism:
    def insert_layer(self, model, layer_type, position):
        """Insert new layer while preserving functionality"""
        new_model = clone_model(model)
        
        if layer_type == 'conv':
            new_layer = Conv2D(filters=32, kernel_size=(3,3))
        elif layer_type == 'pool':
            new_layer = MaxPooling2D(pool_size=(2,2))
            
        # Insert at specified position
        layers = new_model.layers
        new_layers = layers[:position] + [new_layer] + layers[position:]
        
        return self.rebuild_model(new_layers)
    
    def rebuild_model(self, layers):
        """Rebuild model with new architecture"""
        # Implementation details for model reconstruction
        pass

8 Aikace-aikacen Gaba

Ana iya faɗaɗa tsarin da aka tsara zuwa yankuna daban-daban ciki har da bincike na kiwon lafiya, motocin cin gashin kansu, da hasashen kuɗi. Hanyar rarraba tana ba da damar haɓaka samfuri na haɗin gwiwa yayin da ake adana sirrin bayanai ta hanyoyin koyon tarayya.

9 Bincike na Asali

Wannan bincike yana wakiltar babban haɗuwa na fasahar blockchain da koyon inji ta atomatik, yana magance ƙayyadaddun iyakoki a duka fagagen. Yarjejeniyar NAS ɗin da aka tsara ta magance ƙarfin lissafi na binciken tsarin jijiyoyin kwamfuta yayin amfani da hanyoyin ƙarfafawa na blockchain don daidaitawa mara tsari. Wannan hanyar ta yi daidai da manyan abubuwan da suka faru na ƙaddamar da ikon AI, kamar yadda dandamali kamar TensorFlow da PyTorch suka rage shinge ga amfani da koyo mai zurfi.

Tushen fasaha ya ginu akan ingantattun hanyoyin NAS, musamman hanyar morphism na cibiyar sadarwa da aka nuna a cikin AutoKeras, amma yana faɗaɗa shi ta hanyar haɗin gwiwar tushen blockchain. Wannan tsarin rarraba yana magance manyan buƙatun lissafi na algorithms na NAS, wanda bisa ga binciken Google akan EfficientNet na iya buƙatar fiye da kwanaki 1000 na GPU don cikakken binciken tsarin gine-gine. Ta hanyar rarraba wannan aikin a cikin nodes da yawa tare da daidaitaccen ƙarfafawa, tsarin yana rage lokacin bincike yayin da yake kula da ingancin bincike.

Idan aka kwatanta da haɗin gwiwar blockchain-injin koyo kamar OpenMined da SingularityNET, wannan shawara tana mai da hankali musamman kan tsarin ƙirƙira samfuri maimakon raba bayanai ko turawa samfuri. Wannan ƙwarewa yana ba da damar ingantaccen ingantaccen tsarin binciken gine-gine. Yankin hasashen kudi na sirri yana aiki azaman gwajin da ya dace saboda yanayinsa mai rikitarwa, mara tsayi da kuma samun wadataccen bayanan blockchain, kodayake hanyar ta bayyana ta zama gama gari ga wasu matsalolin hasashe na lokaci-lokaci.

Haɗin ingantaccen Bayesian tare da morphism na cibiyar sadarwa yana ba da fa'idodi na ka'idar fiye da hanyoyin ƙarfafa koyo, kamar yadda aka nuna a cikin ainihin aiwatarwar AutoKeras. Tsarin Bayesian yana inganta yanayin aikin tsarin gine-gine yadda ya kamata, yana jagorantar bincike zuwa yankuna masu ban sha'awa yayin guje wa cikakken kimanta duk yuwuwar. Wannan yana da mahimmanci musamman a cikin saitunan rarraba inda dole ne a rage yawan sadarwa tsakanin nodes.

Ci gaban gaba zai iya haɗa ingantaccen manufa da yawa la'akari da daidaito kawai amma har ma da girman samfuri, saurin ƙaddamarwa, da ingantaccen makamashi - mahimman abubuwan da aka yi la'akari da su don turawa a duniyar gaske. Hanyar tana nuna alamar ƙirƙirar hanyoyin koyon inji masu sauƙi da inganci, kodayake ƙalubale game da tabbatar da sakamako da hana wasa na tsarin ƙarfafawa suna buƙatar ƙarin bincike.

10 Nassoshi

1. Zoph, B., & Le, Q. V. (2017). Binciken Tsarin Jijiyoyin Kwamfuta tare da Ƙarfafa Koyo. arXiv:1611.01578
2. Liu, H., Simonyan, K., & Yang, Y. (2019). DARTS: Binciken Tsarin Gine-gine Daban-daban. arXiv:1806.09055
3. Jin, H., Song, Q., & Hu, X. (2019). Auto-Keras: Tsarin Bincike mai Ingantaccen Tsarin Jijiyoyin Kwamfuta. KDD 2019
4. Nakamoto, S. (2008). Bitcoin: Tsarin Kuɗin Lantarki na Peer-to-Peer
5. OpenMined (2020). Koyon inji mai rarraba kiyaye sirri
6. SingularityNET (2019). Kasuwar sabis na AI mara tsari
7. Zhu et al. (2017). Fassarar Hotuna-zuwa-Hoto mara biyu ta amfani da Cibiyoyin Sadarwa masu Juyayin Zagayowar. ICCV 2017
8. Brown et al. (2020). Samfuran Harshe Ƙwararrun Masu Koyo. NeurIPS 2020