综述性文章，但缺乏对学术论文特别是2019年新想法的阐述
https://mattturck.com/ai-blockchain/
https://www.bilibili.com/video/av17379157/
AI and Blockchain: A Disruptive Integration https://www.computer.org/csdl/magazine/co/2018/09/mco2018090048/146z4GiPEHP
Online Deep Reinforcement Learning for Computation Offloading in Blockchain-Empowered Mobile Edge Computing
里面提到的Fetch.ai是比较有意思的，Fetch is a decentralised digital representation of the world in which autonomous software agents perform useful economic work. By bringing data to life, Fetch solves one of the greatest problems in the data industry today: data can't sell itself. With Fetch, it can: data is able to actively take advantage of any opportunity to exploit itself in the marketplace in an environment that's constantly reorganising to make that task as easy as possible.
https://btcmanager.com/cryptocurrency-projects-that-bring-artificial-intelligence-to-the-blockchain/?q=/cryptocurrency-projects-that-bring-artificial-intelligence-to-the-blockchain/&
Blockchain and Artificial Intelligence equals Super untouchable AI https://cryptoporridge.com/2018/07/29/what-could-an-artificially-intelligent-blockchain-do/
What are the differences between Blockchain and Artificial Intelligence?
The main difference between the two is that AI is for extracting answers and forecast from huge amount of data, while the Blockchain is used to verify various types of transactions; so in other words, AI is probabilistic while Blockchain is deterministic.
AI is always changing, while Blockchain is permanent.
AI is algorithm to guess or predict reality(联想？), while Blockchain keeps a record of reality.（记忆？LSTM?）
AI is centralized, while Blockchain is decentralized.
AI has known owners and users, while Blockchain can be anonymous.
What are the similarities between Blockchain and Artificial Intelligence?
Both AI and Blockchain are digital algorithms.
AI and Blockchain are both heavily funded.
AI and Blockchain both attract the best and brightest developers.
https://www.openmined.org/
The OpenMined community is on a mission to create an accessible ecosystem of tools for private, secure, multi-owner governed AI. They will do this by extending popular libraries like TensorFlow and PyTorch with advanced techniques in cryptography and private machine learning.
Machine Learning in/for Blockchain: Future and Challenges
https://www.chainintel.com/ Decentralized and Distributed Artificial Intelligence
ChainIntel is a smart decentralized application (DApps) development platform that facilitates deployment and usage of AI models in DApps.可惜代码很久没有更新了
It augments and distributes the execution of the AI model to different parts of the network, enabling fast, scalable and robust smart DApps.
SingularityNET A Decentralized, Open Market and Network for AIs
Blockchain for AI: Review and Open Research Challenges
the underlying technology of modern digital currency that provides a peer to peer network enabling sensitive participants to communicate and collaborate in a secure way without a fully trusted third party or an authorized central node.
Byzantine attack tolerant machine learning algorithms
There are two main challenges for such a decentralized distributed machine learning: one is how to protect data holders'privacy, and the other is how to guarantee the resilience of the system to malicious users ’attacks, i.e., Byzantine attacks
The decentralized AI enables to process and perform analytics or decision making on trusted, digitally signed, and secure shared data that has been transacted and stored on the blockchain, in a distributed and decentralized fashion, without Trusted Third Parties or intermediaries
近期感兴趣的是AI(数据、模型、算力)以及TEE和区块链的深度结合，而且不是用AI去分析学习预测区块链数据或者币价，而是怎么AI作为生产力，区块链作为生产关系，怎么把区块链的去中心化经济系统激励用在构筑一个去中心化的深度强化学习网络，当然也同时实现联邦学习这样的隐私保护且安全的分布式机器学习，数据也是去中心化存储的并且DID为用户所有-数据会通过bot来销售自己，持续的去中心化的价值互联网的市场交易和撮合过程就是深度学习中的参数和超参数以及连接结构的迭代优化过程，只有构建去中心化的数据、模型、算力市场也不是我兴趣所在
问题：与第一作者电子邮件联系未回，感觉这篇文章有问题，并无实际代码实现，数据是伪造的，没有对本地梯度计算贡献的激励，只有对出块的激励，不过出块节点同时完成全局梯度计算，机器学习算法的去中心化来自全局梯度计算节点的POW;此外就是没有公开代码，不清楚哪些是在智能合约里完成的，能不能在出块时间内完成一次学习，另外要不要修改以太坊代码，目前我理解AI+Blockchain是左右脑
When Machine Learning Meets Blockchain: A Decentralized, Privacy preserving and Secure Design: LearningChain
by considering a general (linear or nonlinear) learning model and without a trusted central server. Specififically, we design a decentralized Stochastic Gradient Descent (SGD) algorithm to learn a general predictive model over the blockchain. In decentralized SGD, we develop differential privacy based schemes to protect each party’s data privacy, and propose an l-nearest aggregation algorithm(差分隐私) to protect the system from potential Byzantine attacks. We also conduct theoretical analysis on the privacy and security of the proposed LearningChain. Finally, we implement LearningChain on Etheurum and demonstrate its effificiency and effectiveness through extensive experiments.
Local Gradient Computation: The data holders fifirstcreate pseudo-identities and calculate the local gradients based on the current global model. Then, they employ a differential privacy scheme to perturb their local gradients. Encapsulated with other related information, the messages are broadcasted to all the computing nodes.
Global Gradient Aggregation: The computing nodes compete to obtain the authority of appending a new block to the chain by solving a mathematical puzzle. Once one computing node wins the game, it applies a Byzantine attack tolerant aggregation scheme to the received local gradients in its memory pool and calculates the global gradient to update the model parameters w. At last, it appends a newly created block with related information to the chain.
Blockchain-Based Privacy Preserving Deep Learning
secure the share of updates in federated learning, Along with the digital signature of a node, a transaction broadcasts to other nodes information including changes of hyperparameters and weights, public keys (participants’addresses). Other nodes validate the transaction and test updates according to their local data sets. If most nodes confirm that the performance score of the updated model is higher than the existing model under their local data sets, the updates are implemented into the existing model.
DeepChain: Auditable and Privacy-Preserving Deep Learning with Blockchain-based Incentive
Deep learning technology has achieved the high-accuracy of state-of-the-art algorithms in a variety of AI tasks. Its popularity has drawn security researchers’ attention to the topic of privacy-preserving deep learning, in which neither training data nor model is expected to be exposed. Recently, federated learning becomes promising for the development of deep learning where multi-parties upload local gradients and a server updates parameters with collected gradients, the privacy issues of which have been discussed widely. In this paper, we explore additional security issues in this case, not merely the privacy. First, we consider that the general assumption of honest-but-curious server is problematic, and the malicious server may break privacy. Second, the malicious server or participants may damage the correctness of training, such as incorrect gradient collecting or parameter updating. Third, we discover that federated learning lacks an effective incentive mechanism for distrustful participants due to privacy and financial considerations. To address the aforementioned issues, we introduce a value-driven incentive mechanism based on Blockchain. Adapted to this incentive setting, we migrate the malicious threats from server and participants, and guarantee the privacy and auditability. Thus, we propose to present DeepChain which gives mistrustful parties incentives to participate in privacy-preserving learning, share gradients and update parameters correctly, and eventually accomplish iterative learning with a win-win result. At last, we give an implementation prototype by integrating deep learning module with a Blockchain development platform (Corda V3.0). We evaluate it in terms of encryption performance and training accuracy, which demonstrates the feasibility of DeepChain
Each party trains her local model independently, and at the end of a local iteration the party generates a contract to trade by attaching her local gradients to the contract.When a party got her local gradients, she sends out the gradients through a smart contract called trading contractto DeepChain.Those contracts can be downloaded to process by worker
Collaborative model training:Parties of a cooperative group train a deep learning model collaboratively. Specifically, after deciding a same deep learning model and parameter initialization, the model is trained in an iterative manner. In each iteration, all parties trades their gradients, and workers download the contracts to process the gradients. The processed gradients are then sent out by workers through smart contract calledprocessing contract. The correctly processed gradients are used to update parameters of the collaborative model by the leader who is selected from workers. Parties download the updated parameters of the collaborative model and update their local models accordingly. After that parties begin next iteration of model training
A Corda network contains multiple notaries, and our consensus protocol introduced in section 4.2.5 can be executed on them. We build nodes and divide them into parties and workers. Specifically, we set up two CorDapps which agree on the Blockchain. The nodes of one CorDapp serve as parties, and the nodes of another one play the role of workers. we implement the above building blocks into three modules, i.e., CordaDeepChain, TrainAlgorithm, and CryptoSystem.
BlockDeepNet: A Blockchain-Based Secure Deep Learning for IoT Network
Blockchain-based edge computing for deep neural network applications
Converging blockchain and next generation artificial intelligence technologies to decentralize and accelerate biomedical research and healthcare
ModelChain: Decentralized Privacy Preserving Healthcare Predictive Modeling Framework on Private Blockchain Networks
Decentralized & Collaborative AI on Blockchain
有代码在https://github.com/microsoft/0xDeCA10B
It then accepts “add data” actions from participants, with the incentive mechanism possibly triggering payments or allowing other actions. Adding data involves validation from the incentive mechanism, storing in the data handler, and fifinally calling the update method on the model’s contract, as shown in Fig. 1. Prediction is done off-chain bycalling the predict function provided for convenience in the model’s smart contract code.
Significant advances are being made in artificial intelligence, but accessing and taking advantage of the machine learning systems making these developments possible can be challenging, especially for those with limited resources. These systems tend to be highly centralized, their predictions are often sold on a per-query basis, and the datasets required to train them are generally proprietary and expensive to create on their own. Additionally, published models run the risk of becoming outdated if new data isn’t regularly provided to retrain them.
We envision a slightly different paradigm, one in which people will be able to easily and cost-effectively run machine learning models with technology they already have, such as browsers and apps on their phones and other devices. In the spirit of democratizing AI, we’re introducing Decentralized & Collaborative AI on Blockchain.
presented at the second IEEE International Conference on Blockchain July 14–17
https://www.microsoft.com/en-us/research/blog/leveraging-blockchain-to-make-machine-learning-models-more-accessible/
propose a framework for participants to collaboratively build a dataset and use smart contracts to host a continuously updated model 一个激励和去中心化的社区驱动下的开源持续学习的框架
我也希望最终将所有机器学习算法用区块链智能合约重新实现一遍，都成为DApp
https://github.com/steerapi/daimon
DaiMoN: A Decentralized Artificial Intelligence Model Network
We introduce DaiMoN, a decentralized artificial intelligence model network, which incentivizes peer collaboration in improving the accuracy of machine learning models for a given classification problem. It is an autonomous network where peers may submit models with improved accuracy and other peers may verify the accuracy improvement. The system maintains an append-only decentralized ledger to keep the log of critical information, including who has trained the model and improved its accuracy, when it has been improved, by how much it has improved, and where to find the newly updated model. DaiMoN rewards these contributing peers with cryptographic tokens. A main feature of DaiMoN is that it allows peers to verify the accuracy improvement of submitted models without knowing the test labels. This is an essential component in order to mitigate intentional model overfitting by model-improving peers. To enable this model accuracy evaluation with hidden test labels, DaiMoN uses a novel learnable \textit{Distance Embedding for Labels} (DEL) function proposed in this paper. Specific to each test dataset, DEL scrambles the test label vector while preserving the distance between the dataset's test label vector and a label vector inferred by the classifier. It therefore allows \textit{proof-of-improvement} (PoI) by peers without knowing any test labels. We provide analysis and empirical evidence that under DEL, peers can accurately assess model accuracy. We also argue that it is hard to invert the embedding function and thus, DEL is resilient against attacks aiming to recover test labels in order to cheat.
Distance Embedding for Labels (DEL), a key technique by which DaiMoN can allow peers to verify the accuracy of a submitted model without knowing the labels of the test dataset.
Trustless Machine Learning Contracts; Evaluating and Exchanging Machine Learning Models on the Ethereum Blockchain
Algorithmia Research - Besir Kurtulmus, Kenny Daniel - 2018在智能合约里实现神经网络推理
Using blockchain technology, it is possible to create contracts that offer a reward in exchange for a trained machine learning model for a particular data set. This would allow users to train machine learning models for a reward in a trustless manner.
The smart contract will use the blockchain to automatically validate the solution, so there would be no debate about whether the solution was correct or not. Users who submit the solutions won’t have counterparty risk that they won’t get paid for their work. Contracts can be created easily by anyone with a dataset, even programmatically by software agents.
https://github.com/algorithmiaio/danku/blob/master/contracts/Danku.sol
function model_accuracy(uint submission_index, int256[datapoint_size][] data) public constant returns (int256){
function get_prediction(Submission sub, int[datapoint_size] data_point)privatepurereturns(int256[]) {
uint[]memoryl_nn =newuint[](sub.num_neurons_hidden_layer.length+2);
l_nn[0]=sub.num_neurons_input_layer;
for(uinti =0; i l_nn[i+1]=sub.num_neurons_hidden_layer[i]; } l_nn[sub.num_neurons_hidden_layer.length+1]=sub.num_neurons_output_layer; returnforward_pass(data_point, sub.weights, sub.biases, l_nn); }