Abstract: A combined hyperparameter and proxy model tuning method is described. The method involves multiple search iterations. In each search iteration, candidate hyperparameters are considered. An initial (‘seed’) hyperparameter is determined, and used to train one or more first proxy models on a target dataset. From the first proxy model(s), one or more first synthetic datasets are sampled. A first evaluation model is fitted to each first synthetic dataset, for each candidate hyperparameter, enabling each candidate hyperparameter to be scored.

Abstract: A multi-level compression method for picture data includes: obtaining original data by performing modulo operation on a picture to-be-compressed; obtaining a first storage amount and a second storage amount by respectively storing the original data in a first container and a second container and calculating a size of data in the first container and a size of data in the second container; determining a bit length of index data, and constructing a color table and an initial index table; obtaining a target color table by identifying a minimum value in the color table and subtracting the minimum value from each data in the color table; obtaining a first compressed index table by compressing the initial index table according to the bit length; and obtaining a second compressed index table by identifying a locator in the first compressed index table and compressing the first compressed index table according to the locator.

Abstract: Disclosed are a testing system and method under fluid-solid coupling effects. A dynamic stress field during fluid flow, deformation of a porous rock framework, real-time evolution of the stress field and deformation of a fluid-solid interface under fluid-solid coupling effects can be obtained, on the basis of a collected photo-elastic stripe image and a surface deformation image. A stress field of a solid framework and fluid in porous rock, and a strain field of the solid framework and the fluid-solid interface under fluid-solid coupling effects can be visually and quantitatively displayed by means of a display device.

Abstract: A method comprising: sampling a temporal causal graph from a temporal graph distribution specifying probabilities of directed causal edges between different variables of a feature vector at a present time step, and from one variable at a preceding time step to another variables at the present time step. Based on this there are identified: a present parent which is a cause of the selected variable in the present time step, and a preceding parent which is a cause of the selected variable from the preceding time step. The method then comprises: inputting a value of each identified present and preceding parent into a respective encoder, resulting in a respective embedding of each of the present and preceding parents; combining the embeddings of the present and preceding parents, resulting in a combined embedding; inputting the combined embedding into a decoder, resulting in a reconstructed value of the selected variable.

Abstract: The disclosure discloses a scheduling method for collaborative work of multi-device in a parallel switching system, the method comprises: providing a load and N backup energy storage devices, n is a positive integer greater than or equal to 2; and connecting the AC output interfaces of the N backup energy storage devices to the load or assembling the AC charging interfaces of the N backup energy storage devices to form a parallel switching system and obtaining numbers 1 to N according to the sequence of the communication modules accessing Wi-Fi hotspots. When the backup energy storage device with any number is switched to the load in the power supply mode, the other backup energy storage devices are switched to the wire mode to transfer current.

Abstract: A method of training a model comprising a generative network mapping a latent vector to a feature vector, wherein weights in the generative network are modelled as probabilistic distributions. The method comprises: a) obtaining one or more observed data points, each comprising an incomplete observation of the features in the feature vector; b) training the model based on the observed data points to learn values of the weights of the generative network which map the latent vector to the feature vector; c) from amongst a plurality of potential next features to observe, searching for a target feature of the feature vector which maximizes a measure of expected reduction in uncertainty in a distribution of said weights of the generative network given the observed data points so far; and d) outputting a request to collect a target data point comprising at least the target feature.

Abstract: Disclosed are a testing system and method under fluid-solid coupling effects. A dynamic stress field during fluid flow, deformation of a porous rock framework, real-time evolution of the stress field and deformation of a fluid-solid interface under fluid-solid coupling effects can be obtained, on the basis of a collected photo-elastic stripe image and a surface deformation image. A stress field of a solid framework and fluid in porous rock, and a strain field of the solid framework and the fluid-solid interface under fluid-solid coupling effects can be visually and quantitatively displayed by means of a display device.

Abstract: The disclosure discloses a scheduling method for collaborative work of multi-device in a parallel switching system, the method comprises:providing a load and N backup energy storage devices, n is a positive integer greater than or equal to 2; and connecting the AC output interfaces of the N backup energy storage devices to the load or assembling the AC charging interfaces of the N backup energy storage devices to form a parallel switching system and obtaining numbers 1 to N according to the sequence of the communication modules accessing Wi-Fi hotspots. When the backup energy storage device with any number is switched to the load in the power supply mode, the other backup energy storage devices are switched to the wire mode to transfer current.

Abstract: A method of training a model comprising a generative network mapping a latent vector to a feature vector, wherein weights in the generative network are modelled as probabilistic distributions. The method comprises: a) obtaining one or more observed data points, each comprising an incomplete observation of the features in the feature vector; b) training the model based on the observed data points to learn values of the weights of the generative network which map the latent vector to the feature vector; c) from amongst a plurality of potential next features to observe, searching for a target feature of the feature vector which maximizes a measure of expected reduction in uncertainty in a distribution of said weights of the generative network given the observed data points so far; and d) outputting a request to collect a target data point comprising at least the target feature.

Abstract: An experiment system and transparent experiment method for replicating fluid displacement in a pore structure of a natural rock mass are provided. The natural pore structure is extracted and a digital porous model corresponding to the natural rock mass is reconstructed with the image processing method. Based on the digital porous model, a three-dimensional pore structure model with a transparent and visible internal structure is printed by a 3D printing device, such that the pore space inside the three-dimensional pore structure model is visible. In this way, the whole fluid flow during the displacement-seepage process within the natural rock mass can be replicated and visually observed from the outside when performing the displacement-seepage experiment. Further, temperature, flow rate, and pressure can be accurately controlled, to replicate various experiment conditions, so as to perform quantitative analysis on distribution features of a seepage field and a fluid speed field.

Abstract: An experiment system and transparent experiment method for replicating fluid displacement in a pore structure of a natural rock mass are provided. The natural pore structure is extracted and a digital porous model corresponding to the natural rock mass is reconstructed with the image processing method. Based on the digital porous model, a three-dimensional pore structure model with a transparent and visible internal structure is printed by a 3D printing device, such that the pore space inside the three-dimensional pore structure model is visible. In this way, the whole fluid flow during the displacement-seepage process within the natural rock mass can be replicated and visually observed from the outside when performing the displacement-seepage experiment. Further, temperature, flow rate, and pressure can be accurately controlled, to replicate various experiment conditions, so as to perform quantitative analysis on distribution features of a seepage field and a fluid speed field.

Abstract: A method of training a model comprising a generative network mapping a latent vector to a feature vector, wherein weights in the generative network are modelled as probabilistic distributions. The method comprises: a) obtaining one or more observed data points, each comprising an incomplete observation of the features in the feature vector; b) training the model based on the observed data points to learn values of the weights of the generative network which map the latent vector to the feature vector; c) from amongst a plurality of potential next features to observe, searching for a target feature of the feature vector which maximizes a measure of expected reduction in uncertainty in a distribution of said weights of the generative network given the observed data points so far; and d) outputting a request to collect a target data point comprising at least the target feature.

Abstract: In various examples there is a data retrieval apparatus. The apparatus has a processor configured to receive a data retrieval request associated with a user. The apparatus also has a machine learning system configured to compute an affinity matrix of users for data items. The affinity matrix has a plurality of observed ratings of data items, and a plurality of predicted ratings of data items. The processor is configured to output a ranked list of data items for the user according to contents of the affinity matrix.