Abstract: A technique capable of providing a new function usable as an activation function is provided. An inference apparatus includes a receiving unit that receives input of target data; and an inference unit that executes a predetermined inference process with respect to the target data using a neural network model. The neural network model includes a plurality of processing layers, and, as the processing layers, one or more activation function layers that convert an input value by a predetermined activation function. The activation function of at least one of the activation function layers is configured as a function of a waveform the output value of which changes, in a first range, to approach a maximum value as an input value increases and, in a second range, away from a minimum value as the input value increases, such that the output values in the first and second ranges are not the same.

Abstract: A processing apparatus includes at least one processor configured to function as: an input unit that receives encrypted data based on homomorphic encryption; and a process execution unit that executes a predetermined process by using the encrypted data while maintaining a secret state and includes one or more processing units. At least one of the processing units is a multiplication corresponding processing unit for executing a calculation corresponding to a processing of multiplying plaintext data by a predetermined multiplier.

Abstract: An inference apparatus that executes a predetermined inference process using a neural network model including: an input unit configured to receive encryption data; and a processor configured to execute a process including: executing predetermined process with respect to the encryption data using the neural network model, wherein the neural network model has a processing layer that executes a process of generating second encryption data from a plurality of pieces of first encryption data input from a previous stage and outputting the second encryption data, the second encryption data providing, by being decrypted, a processing result obtained by performing operations with modulo operation with respect to plurality of pieces of plain text data corresponding to the plurality of pieces of first encryption data, the neural network model being configured such that a nonlinear function is not applied to the second encryption data in a posterior stage of the processing layer.

Abstract: A processing apparatus includes at least one processor configured to function as: an input unit that receives encrypted data based on homomorphic encryption as an input; and a process execution unit that executes a predetermined process by using the encrypted data while maintaining a secret state by encryption and includes one or more processing units. At least one of the processing units is a multiplication corresponding processing unit for executing a calculation in a ciphertext space corresponding to a processing of multiplying plaintext data by a predetermined multiplier. The multiplication corresponding processing unit executes a calculation in the ciphertext space corresponding to a calculation of multiplying the plaintext data by an adjustment multiplication value on first encrypted data input from a preceding stage and outputs resulting data.

Abstract: A processing apparatus includes at least one processor configured to function as: an input unit that receives encrypted data based on homomorphic encryption; and a process execution unit that executes a predetermined process by using the encrypted data while maintaining a secret state and includes one or more processing units. At least one of the processing units is a multiplication corresponding processing unit for executing a calculation corresponding to a processing of multiplying plaintext data by a predetermined multiplier.

Abstract: A processing apparatus includes at least one processor configured to function as: an input unit that receives encrypted data based on homomorphic encryption as an input; and a process execution unit that executes a predetermined process by using the encrypted data while maintaining a secret state by encryption and includes one or more processing units. At least one of the processing units is a multiplication corresponding processing unit for executing a calculation in a ciphertext space corresponding to a processing of multiplying plaintext data by a predetermined multiplier. The multiplication corresponding processing unit executes a calculation in the ciphertext space corresponding to a calculation of multiplying the plaintext data by an adjustment multiplication value on first encrypted data input from a preceding stage and outputs resulting data.

Abstract: An inference apparatus that executes a predetermined inference process using a neural network model including: an input unit configured to receive encryption data; and a processor configured to execute a process including: executing predetermined process with respect to the encryption data using the neural network model, wherein the neural network model has a processing layer that executes a process of generating second encryption data from a plurality of pieces of first encryption data input from a previous stage and outputting the second encryption data, the second encryption data providing, by being decrypted, a processing result obtained by performing operations with modulo operation with respect to plurality of pieces of plain text data corresponding to the plurality of pieces of first encryption data, the neural network model being configured such that a nonlinear function is not applied to the second encryption data in a posterior stage of the processing layer.

Abstract: A technique capable of providing a new function usable as an activation function is provided. An inference apparatus includes a receiving unit that receives input of target data; and an inference unit that executes a predetermined inference process with respect to the target data using a neural network model. The neural network model includes a plurality of processing layers, and, as the processing layers, one or more activation function layers that convert an input value by a predetermined activation function. The activation function of at least one of the activation function layers is configured as a function of a waveform the output value of which changes, in a first range, to approach a maximum value as an input value increases and, in a second range, away from a minimum value as the input value increases, such that the output values in the first and second ranges are not the same.

Abstract: An inference device includes a processor configured to execute a process including: acquiring of a learned model in which a parameter is adjusted by using a first neural network employing a nonlinear function as an activation function, the parameter including at least one of a weight and a bias of coupling between neurons included in the first neural network; setting of a parameter in a second neural network employing an approximation polynomial of the nonlinear function as an activation function in accordance with the learned model; and performing of inference processing on encrypted data as encrypted by using the second neural network in response to the encrypted data being input.