Abstract: Embodiments of the present disclosure provide a solution for image encoding learning and application. A method for image encoding learning comprises: extracting an image feature representation of a sample image using an image encoder to be trained; extracting a text feature representation of a sample text sequence using a text encoder, the sample text sequency being associated with the sample image; generating, using the text encoder, a predicted text sequence based on the text feature representation and the image feature representation; and training the image encoder at least based on a text error between the predicted text sequence and the sample text sequence.

Abstract: The present disclosure discloses a method and an apparatus for generating a vector representation of a knowledge graph, and relates to a field of a field of artificial intelligence technologies. The detailed implementing solution is: obtaining a knowledge graph, the knowledge graph including a plurality of entity nodes; obtaining a context type and context data corresponding to the knowledge graph; and generating vector representations corresponding to the plurality of entity nodes by a context model based on the context data and the context type.

Abstract: There are provided method, apparatus, device, and medium for determining update gradient for contrastive learning model. In the method, a gradient factor of a first type for the contrastive learning model is determined based on a first group of training data and a second group of training data for training the contrastive learning model. The gradient factor of the first type is not used for backpropagation during a training process. In a first stage of the training process, a gradient factor of a second type associated with the first group of training data is determined based on the contrastive learning model. The gradient factor of the second type is used for backpropagation during the training process. Gradient is obtained for updating the contrastive learning model based on the gradient factor of the first type and the gradient factor of the second type associated with the first group of training data.

Abstract: A method of training and applying contrastive learning model. The method includes obtaining a sample set and label information for training contrastive learning model, the sample set including a plurality of first samples of a first modality and a plurality of second samples of a second modality, the label information indicating a correlation between samples of the plurality of first samples and samples of the plurality of second samples; determining whether sample mixing is to be performed on the first modality or the second modality; in accordance with a determination that sample mixing is to be performed on the first modality, generating at least one first mixed sample of the first modality by mixing at least one pair of first samples among the plurality of first samples; and training the contrastive learning model at least based on the at least one first mixed sample and first mixed label information.

Abstract: Methods, apparatuses, a device, and a medium for training a contrastive learning model are provided. In a method, a plurality of sample sets for training the contrastive learning model are obtained, and the plurality of sample sets comprises a first sample set and a second sample set. A first target sample set is selected from the first sample set and the second sample set according to a predetermined rule. A first set of samples are determined based on the first target sample set according to a predefined batch size. The contrastive learning model is trained using the first set of samples. In this way, on the one hand, performance degradation of the contrastive learning model due to sample set bias may be avoided; on the other hand, a forgetting problem in the training process may be alleviated.

Abstract: A method of contrastive learning comprises: determining, based on a model construction criterion, a first encoder for a first modality and a second encoder for a second modality; constructing a first contrastive learning model, the first contrastive learning model comprising the first encoder and a third encoder for the second modality, and a model capacity of the third encoder being greater than a model capacity of the second encoder; performing pre-training of the first contrastive learning model based on a first training dataset for the first modality and the second modality; and providing the pre-trained first encoder in the pre-trained first contrastive learning model for a downstream task. Because only the model capacity of one encoder is increased in the pre-training stage, model performance may be improved without increasing model training overhead during downstream task fine-tuning and model running overhead during model application.

Abstract: A device includes a touch-mode biometric sensor having a first side facing toward a user and a second side opposite to the first side, and a display arranged under the touch-mode biometric sensor and adjacent to the second side and configured to display an image in response to a sensing result, associated with a biometric feature of the user, of the touch-mode biometric sensor.

Abstract: Provided are a data processing system and method based on dynamic redundancy heterogeneous encoding, and a device. The method comprises: respectively performing error correction encoding on information to be processed and a processing rule, so as to form encoded information to be processed and an encoded processing rule; processing, by using the encoded processing rule, the encoded information to be processed, so as to obtain response data; and then performing error correction decoding on N pieces of response data, so as to obtain processing result information of the information to be processed.

Abstract: Provided in the present application are a data processing method and apparatus based on data coding, and a device. The method comprises: performing N-channel error correction coding on data to be processed that is in information to be processed, so as to obtain N pieces of coded data to be processed; by using coded meta-channel data obtained by means of error correction coding, performing redundancy processing on the coded data to be processed, so as to obtain N pieces of response data; and then performing error correction decoding on the N pieces of response data, so as to obtain processing result information of the information to be processed.

Abstract: The present application provides an instruction encoding-based data processing method and apparatus, and a device. N-path error correction encoding is performed on an instruction in information to be processed to obtain N encoded instructions, the encoded instructions and encoded meta channel programs obtained by the error correction encoding are used to perform redundant processing on data to be processed to obtain N pieces of response data, and then error correction decoding is performed on the N pieces of response data to obtain processing result information of the information to be processed. Because the N encoded instructions are heterogeneous and the encoded meta channel programs used in the N-path processing are heterogeneous, the randomness of the processing process can be improved, generalized disturbance in the data processing process can be corrected in combination with the error correction encoding and decoding methods, and thus the security of data processing is improved.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech synthesis. The methods, systems, and apparatus include actions of obtaining an audio representation of speech of a target speaker, obtaining input text for which speech is to be synthesized in a voice of the target speaker, generating a speaker vector by providing the audio representation to a speaker encoder engine that is trained to distinguish speakers from one another, generating an audio representation of the input text spoken in the voice of the target speaker by providing the input text and the speaker vector to a spectrogram generation engine that is trained using voices of reference speakers to generate audio representations, and providing the audio representation of the input text spoken in the voice of the target speaker for output.

Abstract: A thermoplastic polyester elastomer conjugate fiber including a core and a sheath is provided. A volume ratio of the core to the sheath is in a range of 4:6 to 6:4.

Abstract: A visible light shielding structure is provided. The visible light shielding structure includes a metal foil layer, two adhesive layers, and two synthesized paper layers. The metal foil layer has a first surface and a second surface opposite to the first surface. The two adhesive layers are disposed on the first surface and the second surface of the metal foil layer, respectively. The two synthesized paper layers are disposed on the two adhesive layers, respectively. Each of the two synthesized paper layers is a polypropylene based resin layer.

Abstract: A release film for high-capacity multilayer ceramic capacitor (MLCC) 0.5-0.8 ?m ceramic slurry and a production method thereof. The release film includes at least three layers of co-extruded polyester resin having a layer A, a layer B and a layer C, and stretching the same in longitudinal and transverse directions. In addition, the layer B includes recycled polyester resin, the layers A and C include new polyester resin raw materials, and a releasing agent having a phenyl releasing control agent. The production method for the release film includes coating a releasing agent on one or both sides of a polyester film, and winding into a roll after drying and hardening.

Abstract: An eco-friendly synthetic paper and a method for manufacturing the same are provided. The eco-friendly synthetic paper includes a light-shielding adhesive layer and two polypropylene-based resin layers. The light-shielding adhesive layer has a first surface and a second surface opposite to the first surface. The two polypropylene-based resin layers are respectively attached to the first surface and the second surface of the light-shielding adhesive layer. The light-shielding adhesive layer contains an adhesive component and a black colorant. The adhesive component is present in an amount between 35.9 wt % and 63.1 wt % of the total weight of the light-shielding bonding layer, and the black colorant is present in an amount between 12.9 wt % and 21.7 wt % of the total weight of the light-shielding bonding layer.

Abstract: A liquid crystal polymer film, and a composite film of liquid crystal polymer and polyimide and a manufacturing method thereof are provided. The liquid crystal polymer film includes 63 wt % to 74 wt % of p-hydroxybenzoic acid, 21 wt % to 26 wt % of 6-hydroxy-2-naphthoic acid, and 5 wt % to 11 wt % of p-hydroxycinnamic acid. The composite film is manufactured by thermocompressing a single layer or multi layers of liquid crystal polymer film and polyimide film so that the composite film can have high flatness and the surface roughness Sa of the composite film is ranging from 0.1 ?m to 10 ?m. In the production process of the composite film, the composite film is rolled up and attached to a copper foil to form a high frequency substrate with good processability. After peeling the polyimide film, the liquid crystal polymer film can be thermocompressed to form a four-layered, six-layered, eight-layered or eight-layered high frequency substrate.

Abstract: An optical clear adhesive (OCA) and a manufacturing method thereof are provided. The OCA is a three-layer laminated structure including an OCA transparent film, a high release film coated on a lower surface of the OCA transparent film and a light release film coated on an upper surface of the OCA transparent. The high release film and the light release film contain 0.01 to 3 wt % of UV absorber. After UV light irradiation, the OCA transparent film has a suitable peeling force between the light release film and the high release film.

Abstract: A release film for high-capacity multilayer ceramic capacitor (MLCC) 0.5-0.8 ?m ceramic slurry and a production method thereof. The release film includes at least three layers of co-extruded polyester resin having a layer A, a layer B and a layer C, and stretching the same in longitudinal and transverse directions. In addition, the layer B includes recycled polyester resin, the layers A and C include new polyester resin raw materials, and a releasing agent having a phenyl releasing control agent. The production method for the release film includes coating a releasing agent on one or both sides of a polyester film, and winding into a roll after drying and hardening.

Abstract: An electrical connector for electrically connecting a chip module to a circuit board includes a body, having a plurality of accommodating slots penetrating the body, where a slot wall of at least one of the accommodating slots protrudes inward to form a boss, and the boss has a stopping corner; and a plurality of conductive terminals, correspondingly accommodated in the accommodating slots. At least one of the conductive terminals has an inclined portion, formed by tapering one side of the conductive terminal, wherein the inclined portion is adjacent to the stopping corner, the inclined portion is not parallel to a top surface of the boss, and the inclined portion is at least partially located above the stopping corner and configured to limit the conductive terminal from moving downward.