Embodiments of a device include an integrated circuit, a reconfigurable stream switch formed in the integrated circuit along with a plurality of convolution accelerators and an arithmetic unit coupled to the reconfigurable stream switch. The arithmetic unit has at least one input and at least one output. The at least one input is arranged to receive streaming data passed through the reconfigurable stream switch, and the at least one output is arranged to stream resultant data through the reconfigurable stream switch. The arithmetic unit also has a plurality of data paths. At least one of the plurality of data paths is solely dedicated to performance of operations that accelerate an activation function represented in the form of a piece-wise second order polynomial approximation.

In one embodiment, a method for improving cardinality estimation of a join predicate between a fact table and an overloaded dimension table is provided. The method includes receiving a dimension table and a fact table in a join predicate of one or more SQL statements. The method further includes identifying a majority of records in the fact table that refer to a subset of records in the dimension table. The method further includes computing a filter factor of the join predicate between the dimension table and the fact table. The method further includes creating a statistical view using one or more relevant portions of the dimension table that are referred to by the fact table.

A method and apparatus for interpolating data on a data grid having a plurality of data grid points to provide an interpolated data value at an interpolated point offset from a data grid point by offsets x, y, the method comprising the steps of obtaining a data value and a plurality of derivative values for each of a set of data grid points defining a sub-grid, defining a 4×4 data value matrix from the data values and derivative values, defining a plurality of 4×4 transformation matrices, calculating a 4×4 coefficient matrix from the transformation matrices and the data value matrix, defining a x vector based on the offset x and a y vector based on the offset y, and calculating the interpolated data value from the x vector, y vector and coefficient matrix.

One embodiment of the invention disclosed herein provides techniques for fitting a mesh representation that includes a plurality of mesh points with a NURBS surface. A subdividing engine subdivides the mesh representation into a plurality of patches in a parametric domain. A vertex solving engine computes one or more vertex boundary continuity constraints for each vertex included in a plurality of vertices associated with the plurality of patches. An edge solving engine computes one or more edge boundary continuity constraints for each edge included in a plurality of edges associated with the plurality of patches. A patch solving engine fits a first patch included in the plurality of patches with at least one partial NURBS surface based on the vertex boundary continuity constraints and the edge boundary continuity constraints.

A strict reverse navigation method for optimal estimation of fine alignment is provided. The strict reverse navigation method including: establishing an adaptive control function; performing a forward navigation calculation process; performing a reverse navigation calculation process; and performing the adaptive control for a number of forward and reverse calculations. The strict reverse navigation method shortens an alignment time for the optimal estimation of fine alignment while ensuring an alignment accuracy. The strict reverse navigation method provided effectively solves a problem that an error of an initial value of filtering in an initial stage of the optimal estimation of fine alignment affects convergence speeds of subsequent stages. In the initial stage, a larger number of the forward and reverse navigation calculations are adopted to reduce an error of the initial value as much as possible and increase a convergence speed of the filtering.

A digital medium environment is described to dynamically modify or extend an existing path in a user interface. An un-parameterized input is received that is originated by user interaction with a user interface to specify a path to be drawn. A parameterized path is fit as a mathematical ordering representation of the path to be drawn as specified by the un-parametrized input. A determination is made as to whether the parameterized path is to extend or modify the existing path in the user interface. The existing path is modified or extended in the user interface using the parameterized path in response to the determining that the parameterized path is to modify or extend the existing path.

The present disclosure describes a system, method, and computer program for real-time and computationally efficient calculation of a recommended value range for a quote variable, such as price, discount, volume, or closing time. The system uses the highest-density interval (HDI) of probability density function (PDF) as a recommended or suggested value range for a quote variable. PDFs for the quote variable are precomputed for groups of related inputs, and each PDF is summarized as an array of discrete points. A dimension reduction technique is applied to the PDF inputs in both the training and real-time (non-training) phases to reduce the number of possible combinations of PDFs. During a quote-creation process, a PDF look-up table enables the system to efficiently identify an applicable PDF from the group of precomputed PDFs based on reduced input values.

A matrix processing apparatus having a three-dimensional slice access memory and an input-/output block. The slice access memory includes cells organized into cell slices, each slice storing an entire selected data matrix. The three-dimensional slice access memory is configured to allow read/write access to the entire data matrix at the same time. The input/output block is connected to the three-dimensional slice access memory and is configured to format data into a format acceptable to the three-dimensional slice access memory.

A simulation device includes a system model, a data selection processing unit, a plurality of observation models, a post-distribution creating unit, a post-distribution unifying unit, and a determining unit. The system model calculates a time evolution of a state vector. The data selection processing unit selects multiple items of observation data. The observation model converts the state vector from the system model on the basis of the relationship with the observation data. The post-distribution creating unit creates, on the basis of the state vector from the observation model and the selected observation data, a first post-distribution based on all pieces of the observation data or a second post-distribution based on absent observation data. The post-distribution unifying unit unifies the first and second post-distributions. The determining unit determines which of the second post-distribution or the unified post-distribution is to be used.

A data acquisition unit acquires input data and output data of a system to be identified. A memory unit stores initial binary/ternary structure matrices obtained from prior knowledge. A binary/ternary structure matrix generation unit reads the initial binary matrices from the memory unit and generates a plurality of sets of candidate binary matrices. A binary structure matrix selection unit selects one set of binary matrices from the sets of candidate binary matrices. A determination unit determines matrices used in a state equation for the system to be identified and matrices used in an output equation for the system to be identified in response to the selected set of binary matrices. An evaluation unit evaluates whether the determined matrices are reasonable for identifying the system to be identified.