Abstract: A system, method, and computer program product for incorporating knowledge from more complex models in simpler models. A method may include obtaining first training data associated with a first set of features and second training data associated with a second set of features different than the first set of features; training a first model based on the first training data and the second training data; and training a second model, using a loss function that depends on an output of an intermediate layer of the first model and an output of the second model, based on the second training data.

Abstract: A system, method, and computer program product for incorporating knowledge from more complex models in simpler models. A method may include obtaining first training data associated with a first set of features and second training data associated with a second set of features different than the first set of features; training a first model based on the first training data and the second training data; and training a second model, using a loss function that depends on an output of an intermediate layer of the first model and an output of the second model, based on the second training data.

Abstract: Methods and systems are described. A method includes accessing transaction data related to restaurants associated with a plurality of geographically separate locations, determining a number of co-visitors shared by each of the restaurants associated with the plurality of geographically separate locations above a predetermined threshold, generating a graphical representation of the plurality of restaurants based on the number of the co-visitors shared by restaurants with the co-visitors above the predetermined threshold and the distance between the restaurants with the co-visitors. The graphical representation is transformed into restaurant embeddings and a neural network model is used to generate restaurant preferences based on the restaurant embeddings.

Abstract: This invention provides a distributed energy transaction matching method based on energy network constraints and multiple knapsack model. First, it considers the matching between financial market transactions and actual energy network scheduling, then generates security constraints on the results of distributed transactions. Next, it designs a multiple knapsack model for peer-to-peer distributed transactions, which can meet the common quotation needs of users or producers who issue transactions and achieve efficient matching under energy network security checks. Finally, the model is verified based on the blockchain Ethereum smart contract test verification. This method provides new ideas for the connection between distributed energy trading and actual physical dispatch, proposes an inclusive and efficient matching model for the point-to-point distributed trading market, which has great reference value for the connection between distributed energy trading and actual conditions.

Abstract: Methods and systems are described. A method includes accessing transaction data related to restaurants associated with a plurality of geographically separate locations, determining a number of co-visitors shared by each of the restaurants associated with the plurality of geographically separate locations above a predetermined threshold, generating a graphical representation of the plurality of restaurants based on the number of the co-visitors shared by restaurants with the co-visitors above the predetermined threshold and the distance between the restaurants with the co-visitors. The graphical representation is transformed into restaurant embeddings and a neural network model is used to generate restaurant preferences based on the restaurant embeddings.

Abstract: Methods and systems are described. A method includes accessing transaction data related to restaurants associated with a plurality of geographically separate locations, determining a number of co-visitors shared by each of the restaurants associated with the plurality of geographically separate locations above a predetermined threshold, generating a graphical representation of the plurality of restaurants based on the number of the co-visitors shared by restaurants with the co-visitors above the predetermined threshold and the distance between the restaurants with the co-visitors. The graphical representation is transformed into restaurant embeddings and a neural network model is used to generate restaurant preferences based on the restaurant embeddings.

Abstract: This invention provides a method for building a multi-energy trading and management platform based on blockchain. Based on the attributes of different functions to be implemented in the energy trading and management platform, the platform is built by combining different advantages of blockchain technology and database technology. Deploy the functions of issuing transactions, matching transactions, and querying transaction information in Ethereum smart contract, store user registration information in a traditional database, and establish a mapping relationship between real information and lengthy blockchain accounts to make information of transactions on the market be displayed with real names. This method not only meets the needs of energy distributed transactions but also avoids the problem of incompatible string type data in Ethereum smart contracts which affects system performance. At the same time, the transaction data displayed with real names ensures the openness and legality of the transaction system.

Abstract: This invention provides a distributed energy transaction matching method based on energy network constraints and multiple knapsack model. First, it considers the matching between financial market transactions and actual energy network scheduling, then generates security constraints on the results of distributed transactions. Next, it designs a multiple knapsack model for peer-to-peer distributed transactions, which can meet the common quotation needs of users or producers who issue transactions and achieve efficient matching under energy network security checks. Finally, the model is verified based on the blockchain Ethereum smart contract test verification. This method provides new ideas for the connection between distributed energy trading and actual physical dispatch, proposes an inclusive and efficient matching model for the point-to-point distributed trading market, which has great reference value for the connection between distributed energy trading and actual conditions.

Abstract: This invention involves a multi-energy distributed transaction and smart contract design method based on the Ethereum private chain. This method includes the following steps: 1) Setting the genesis node of the smart contract as an administrator node, while other nodes as user nodes; 2) Distinguishing system functions and account permissions between administrator nodes and user nodes. Only administrator nodes can add or delete the types of tradable energy in the blockchain through the energy switch function; 3) The user realizes multiple energy transactions at the same time, and the smart contract matches the information released by users based on the market supply and demand situation. This invention can effectively solve the problems of the centralization of the current energy trading system and insufficient security protection. At the same time, it provides an effective solution to the current limited development of blockchain technology in the field of multi-energy trading.

Abstract: Methods and systems are described. A method includes accessing transaction data related to restaurants associated with a plurality of geographically separate locations, determining a number of co-visitors shared by each of the restaurants associated with the plurality of geographically separate locations above a predetermined threshold, generating a graphical representation of the plurality of restaurants based on the number of the co-visitors shared by restaurants with the co-visitors above the predetermined threshold and the distance between the restaurants with the co-visitors. The graphical representation is transformed into restaurant embeddings and a neural network model is used to generate restaurant preferences based on the restaurant embeddings.

Abstract: A system, method, and computer program product for incorporating knowledge from more complex models in simpler models. A method may include obtaining first training data associated with a first set of features and second training data associated with a second set of features different than the first set of features; training a first model based on the first training data and the second training data; and training a second model, using a loss function that depends on an output of an intermediate layer of the first model and an output of the second model, based on the second training data.