Abstract: One embodiment provides for a graphics processing unit (GPU) to accelerate machine learning operations, the GPU comprising an instruction cache to store a first instruction and a second instruction, the first instruction to cause the GPU to perform a floating-point operation, including a multi-dimensional floating-point operation, and the second instruction to cause the GPU to perform an integer operation; and a general-purpose graphics compute unit having a single instruction, multiple thread (SIMT) architecture, the general-purpose graphics compute unit to simultaneously execute the first instruction and the second instruction, wherein the integer operation corresponds to a memory address calculation.

Abstract: One embodiment provides for a compute apparatus to perform machine learning operations, the apparatus comprising a decode unit to decode a single instruction into a decoded instruction, the decoded instruction to perform one or more machine learning operations, wherein the decode unit, based on parameters of the one or more machine learning operations, is to request a scheduler to schedule the one or more machine learning operations to one of an array of programmable compute units and a fixed function compute unit.

Abstract: In an example, an apparatus comprises a compute engine comprising a high precision component and a low precision component; and logic, at least partially including hardware logic, to receive instructions in the compute engine; select at least one of the high precision component or the low precision component to execute the instructions; and apply a gate to at least one of the high precision component or the low precision component to execute the instructions. Other embodiments are also disclosed and claimed.

Abstract: One embodiment provides for a compute apparatus to perform machine learning operations, the compute apparatus comprising instruction decode logic to decode a single instruction including multiple operands into a single decoded instruction, the multiple operands having differing precisions and a general-purpose graphics compute unit including a first logic unit and a second logic unit, the general-purpose graphics compute unit to execute the single decoded instruction, wherein to execute the single decoded instruction includes to perform a first instruction operation on a first set of operands of the multiple operands at a first precision and a simultaneously perform second instruction operation on a second set of operands of the multiple operands at a second precision.

Abstract: An apparatus to facilitate compute optimization is disclosed. The apparatus includes sorting logic to sort processing threads into thread groups based on bit depth of floating point thread operations.

Abstract: In an example, an apparatus comprises a plurality of execution units comprising at least a first type of execution unit and a second type of execution unit and logic, at least partially including hardware logic, to analyze a workload and assign the workload to one of the first type of execution unit or the second type of execution unit. Other embodiments are also disclosed and claimed.

Abstract: An apparatus to facilitate compute optimization is disclosed. The apparatus includes sorting logic to sort processing threads into thread groups based on bit depth of floating point thread operations.

Abstract: An apparatus to facilitate neural network (NN) training is disclosed. The apparatus includes training logic to receive one or more network constraints and train the NN by automatically determining a best network layout and parameters based on the network constraints.

Abstract: Methods and apparatus relating to autonomous vehicle neural network optimization techniques are described. In an embodiment, the difference between a first training dataset to be used for a neural network and a second training dataset to be used for the neural network is detected. The second training dataset is authenticated in response to the detection of the difference. The neural network is used to assist in an autonomous vehicle/driving. Other embodiments are also disclosed and claimed.

Abstract: An apparatus to facilitate optimization of a convolutional neural network (CNN) is disclosed. The apparatus includes optimization logic to receive a CNN model having a list of instructions and including pruning logic to optimize the list of instructions by eliminating branches in the list of instructions that comprise a weight value of 0.

Abstract: An integrated circuit (IC) package apparatus is disclosed. The IC package includes one or more processing units and a bridge, mounted below the one or more processing unit, including one or more arithmetic logic units (ALUs) to perform atomic operations.

Abstract: A motor vehicle sub-frame is provided. That motor vehicle sub-frame includes a front cross member, a rear cross member and first and second side rails connected to the front and rear cross members at each side of the sub-frame. The side rails both include a first component having a first extrusion axis and a second component having a second extrusion axis. The first and second components are joined together so that the two components effectively form an at least partially closed or boxed construction providing enhanced bending and torsion resistances while minimizing component weight.

Abstract: For generating 3D audio content from a two-channel stereo signal, the stereo signal (x(t)) is partitioned into overlapping sample blocks and is transformed into time-frequency domain. From the stereo signal directional and ambient signal components are separated, wherein the estimated directions of the directional components are changed by a predetermined factor, wherein, if changes are within a predetermined interval, they are combined in order to form a directional centre channel object signal. For the other directions an encoding to Higher Order Ambisonics (HOA) is performed. Additional ambient signal channels are generated by de-correlation and rating by gain factors, followed by encoding to HOA. The directional HOA signals and the ambient HOA signals are combined, and the combined HOA signal and the centre channel object signals are transformed to time domain.

Abstract: The invention relates to generating regional tropospheric correction information for correcting observations useful for estimating phase ambiguities and/or a position of global or regional navigation satellite systems (NSS) receiver(s). For each of a plurality of reference stations, at least one troposphere correction parameter is estimated by evaluating NSS observation equations using i) precise satellite information or the information derived from the precise satellite information, and ii) received multiple-frequency-signals-based raw observations or a linear combination thereof. The regional tropospheric correction information is then generated based on the estimated troposphere correction parameter(s) per reference station, the tropospheric correction information comprising a regional tropospheric delay function(s) and coefficients representing a tropospheric delay affecting a NSS signal passing through the troposphere in a region of interest.

Abstract: Methods and apparatus relating to an adaptive partition mechanism with arbitrary tile shape for tile based rendering GPU (Graphics Processing Unit) architecture are described. In an embodiment, the primitive intersection cost value for each atomic tile of an image are determined at least partially based on a vertex element size, a vertex shader length, and a number of the vertices of a primitive of the image. Other embodiments are also disclosed and claimed.

Abstract: The invention relates to methods, apparatuses and computer programs for generating receiver-specific correction information for correcting pseudorange observations. The method comprises: receiving raw observations obtained by the NSS receiver observing NSS multiple frequency signals from a plurality of NSS satellites over multiple epochs; obtaining precise satellite information on: (i) the orbit position of each of the satellites, (ii) a clock offset of each of the satellites, and (iii) a set of biases associated with each of the satellites; estimating ambiguities in the carrier phase of the received raw observations, using the precise satellite information, or information derived therefrom; computing combination values based on the received raw observations together with the estimated ambiguities, to cancel out the effects of the geometry, the effects of the clocks, troposphere and ionosphere; and generating the correction information per satellite, based on the computed combination values.

Abstract: A challenge of audio watermarking systems in which an acoustic path is involved is the robustness against microphone pickup in case of surrounding noise. The strength of phase-based watermarking is increased by determining a masking threshold for a current frequency bin in a frequency/phase representation changing the phase based on that masking threshold and an allowed phase change value, calculating an allowed magnitude change value for the current frequency bin and calculating from an audio quality level value a magnitude change scaling factor for the magnitude change value, and increasing its magnitude accordingly.

Abstract: As a potential format for next-generation audio, techniques for embedding digital watermarks in the Higher Order Ambisonics (HOA) representation of a sound field have been proposed. The inventive embedding method is adapted for watermarking a two-dimensional or three-dimensional Ambisonics representation of a sound field, wherein the Ambisonics representation is decomposed into directional signals and ambient components and includes estimated dominant directions, and wherein the order of the ambient components can be reduced, and wherein watermark information data are embedded in the directional signals, and at receiver side are regained from the watermarked directional signals.

Abstract: A code is generated for mapping source to target code words which allows encoding source data at reduced probability of incorrect decoding, e.g. for DNA storage. The target code words are grouped into subsets and comprise identifying and remaining portions. The identifying portions of target code words corresponding to a same subset are identical. A first code symbol set of source code words is selected for addressing the subsets. For the subsets, neighboring subsets are determined. The identifying portions of the target code words of neighboring subsets differ from those of the corresponding subset by up to a predetermined amount of symbols. Source code words are assigned where the corresponding first code symbols address the same subset to said subset such that an amount of target code words of said subset having their remaining portions identical to their neighboring subsets corresponds to an optimization criterion.

Abstract: An ospemifene solid dispersion for enhancing solubility and bioavailability, and methods of preparation thereof are disclosed. The solid dispersion comprises an active pharmaceutical ingredient, such as ospemifene, and a hydfophilie carrier, such as copovidone, hypromellose acetate succinate, polyvinylpyrrolidine, a polyvinylpyrrolidine/vinyl acetate co-polymer, hydroxyl propyl methylcellulose, hypromellose acetate succinate, a Eudragit® compound, hydroxypropylcellulose, a polyvinyl caprolactam-polyvinyl acetate polyethylene glycol graft co-polymer, hydroxypropyl methylcellulose phthalate, and mixtures thereof, and optionally a surfactant The ospemifene solid, dispersions can be used in methods of treating a symptom related, to menopause, such as vaginal dryness or sexual dysfunction, or in methods of treating osteoporosis.