Abstract: The present disclosure provides for devices, systems, and methods which optimize throughput of bonded connections over multiple variable bandwidth logical paths by adjusting a tunnel bandwidth weighting schema during a data transfer session in response to a change in bandwidth capabilities of one or more tunnels. By making such adjustments, embodiments of the present invention are able to optimize the bandwidth potential of multiple connections being used in a session, while minimizing the adverse consequences of reduced bandwidth issues which may occur during the data transfer session.

Abstract: A circuit for a multistage sequential data process includes a plurality of memory units. Each memory unit is associated with a stage of the sequential data process which, for each data set inputted to the stage, the stage provides an intermediate data set for storage in the associated memory unit for use in at least one subsequent stage of the sequential data process, where each of the plurality of memory units is sized based on relative locations of the stage providing the intermediate data set and the at least one subsequent stage in the sequential data process.

Abstract: The present disclosure describes a closed-loop fluid resuscitation and/or cardiovascular drug administration system that uses continuous measurements and adaptive control architecture. The adaptive control architecture uses a function approximator to identify unknown dynamics and physiological parameters of a patient to compute appropriate infusion rates and to regulate the endpoint of resuscitation.

Abstract: In one embodiment, a processor is to store a membrane potential of a neural unit of a neural network; and calculate, at a particular time-step of the neural network, a change to the membrane potential of the neural unit occurring over multiple time-steps that have elapsed since the last time-step at which the membrane potential was updated, wherein each of the multiple time-steps that have elapsed since the last time-step is associated with at least one input to the neural unit that affects the membrane potential of the neural unit.

Abstract: A system that includes an edge node in signal communication an interior node. The edge node is configured to receive a first correlithm object from a first device outside of the network, to identify an input correlithm object from a node table with the shortest distance, to fetch a second correlithm object from the node table linked with the identified input correlithm object, and to send the second correlithm object to the interior node. The edge node is further configured to receive a third correlithm object from the interior node in response to sending the second correlithm to the interior node, to identify an input correlithm object from the node table with the shortest distance, to fetch a fourth correlithm object from the node table linked with the identified input correlithm object, and to send the fourth correlithm object to the first device.

Abstract: A system configured to emulate a correlithm object processing system includes an input node, a first output node, and a second output node. The input node receives a real-world numeric value comprising a mantissa value and an exponent value. The first output node receives the mantissa value and generates a first correlithm object associated with the mantissa value. The second output node receives the exponent value and generates a second correlithm object associated with the exponent value. A string correlithm object engine maps the first correlithm object to a first sub-string correlithm object of a string correlithm object, and maps the second correlithm object to a second sub-string correlithm object of the string correlithm object.

Abstract: The present disclosure relates to a processor for implementing artificial neural networks, for example, convolutional neural networks. The processor includes a memory controller group, an on-chip bus and a processor core, wherein the processor core further includes a register map, an instruction module, a data transferring controller, a data writing scheduling unit, a buffer module, a convolution operation unit and a hybrid computation unit. The processor of the present disclosure may be used for implementing various neural networks with increased computation efficiency.

Abstract: The present disclosure provides an operation device, comprising: an operation module for executing a neural network operation; and a power conversion module connected to the operation module, for converting input neuron data and/or output neuron data of the neural network operation into power neuron data. The present disclosure further provides an operation method. The operation device and method according to the present disclosure reduce the cost of storage resources and computing resources and increase the operation speed.

Abstract: A distributed node network to emulate a correlithm object processing system includes a resolution node, a first calculation node, and a second calculation node. The first and second calculation nodes determine n-dimensional distances between an input correlithm object and each of the source correlithm objects that are stored in respective portions of a correlithm object mapping table. The resolution node compares the determined n-dimensional distances from the first calculation node with the determined n-dimensional distance associated with the identified source correlithm object from the second calculation node to identify the smallest determined n-dimensional distance based on the comparison. The resolution node then identifies the target correlithm object associated with the smallest determined n-dimensional distance, and outputs the identified target correlithm object.

Abstract: Systems and methods are disclosed herein for recommending an educational course to a user, and may comprise receiving data records associated with availability of a plurality of educational courses at one or more institutions; receiving educational course data and educational course focus data associated with the user; receiving prior user data records comprising prior user educational course data and prior user educational course focus data; determining index scores for each of the plurality of educational courses based upon a similarity between the educational course data and prior user educational course data, and based upon a similarity between the educational course focus data and prior user educational course focus data; and providing a recommended educational course from the plurality of educational courses to the user based upon the determined index scores.

Abstract: A method, computer program product, and system perform computations using a sparse convolutional neural network accelerator. A first vector comprising only non-zero weight values and first associated positions of the non-zero weight values within a 3D space is received. A second vector comprising only non-zero input activation values and second associated positions of the non-zero input activation values within a 2D space is received. The non-zero weight values are multiplied with the non-zero input activation values, within a multiplier array, to produce a third vector of products. The first associated positions are combined with the second associated positions to produce a fourth vector of positions, where each position in the fourth vector is associated with a respective product in the third vector. The products in the third vector are transmitted to adders in an accumulator array, based on the position associated with each one of the products.

Abstract: A distributed node network to emulate a correlithm object processing system includes a resolution node, a first calculation node, and a second calculation node. The first and second calculation nodes store a plurality of first source correlithm objects and a plurality of second source correlithm objects of a mapping table, respectively. The resolution node adds the n-dimensional distance determined for each of the first source correlithm objects with the n-dimensional distance determined for a corresponding one of the second source correlithm objects such that each pair of first and second source correlithm objects has an aggregate n-dimensional distance calculation. It compares these aggregate n-dimensional distance calculations to identify the pair of first and second source correlithm objects with the smallest aggregate n-dimensional distance to an input correlithm object. It identifies the target correlithm object associated with the identified pair, and outputs that target correlithm object.

Abstract: The present technical disclosure relates to artificial neural networks, e.g., gated recurrent unit (GRU). In particular, the present technical disclosure relates to how to implement a hardware accelerator for compressed GRU based on an embedded FPGA. Specifically, it proposes an overall design processing method of matrix decoding, matrix-vector multiplication, vector accumulation and activation function. In another aspect, the present technical disclosure proposes an overall hardware design to implement and accelerate the above process.

Abstract: A distributed node network to emulate a correlithm object processing system, includes a resolution node, a first calculation node, and a second calculation node. The first and second calculation nodes determine the n-dimensional distances between an input correlithm object and first and second portions of each source correlithm object of a mapping table, respectively. The resolution node adds the n-dimensional distance determined for the first portion of each source correlithm object with the n-dimensional distance determined for the second portion of each source correlithm object. It compares these determined n-dimensional distances to each other to identify the source correlithm object with the smallest n-dimensional distance to the input correlithm object. It identifies the target correlithm object that corresponds to the identified source correlithm object, and outputs that target correlithm object.

Abstract: A method of operating an apparatus using a control system that includes at least one neural network. The method includes receiving an input value captured by the apparatus, processing the input value using the at least one neural network of the control system implemented on first one or more solid-state chips, and obtaining an output from the at least one neural network resulting from processing the input value. The method may also include processing the output with another neural network implemented on solid-state chips to determine whether the output breaches a predetermined condition that is unchangeable after an initial installation onto the control system. The aforementioned another neural network is prevented from being retrained. The method may also include the step of using the output from the at least one neural network to control the apparatus unless the output breaches the predetermined condition. Similar corresponding apparatuses are described.

Abstract: An apparatus having components implemented on one or more solid-state chips. The apparatus includes an input device constructed to generate an input data value (input value), and a neural network implemented on solid-state chips trained to generate an output to control the apparatus by processing the input value. The apparatus also includes another neural network implemented on solid-state chips and configured to receive the output from the neural network. The another neural network is trained to determine whether the output of the neural network corresponds to a predetermined condition and generate a control output from the output of the neural network. The apparatus includes a processor configured receive the control output from the aforementioned another neural network, and in response to the control output indicating the output of the first neural network corresponds to a predetermined condition, and control an operation of the neural network. Corresponding methods are also disclosed.

Abstract: Method of training a neural network, including receiving sets of digital attributes representing multidimensional regression at inputs; expanding the network by adding neurons and defining their activation functions and interconnections; any neuron of the neural network is capable of being directly connected to any other neuron; (i) when a training speed falls below predefined threshold, and accuracy does not improve, identifying a neuron with highest error value; (ii) adding a neuron directly between the identified neuron and a corresponding output; (iii) setting only a connection coefficient between the added neuron and the identified neuron to zero before it is modified by the training, while other coefficients of the added neuron are set the same as coefficients of the identified neuron, before they are modified. After at least one iteration, either (iv) finishing the training of the neural network or (v) continuing to train the network to reach a predefined depth.

Abstract: Some embodiments of the invention provide an integrated circuit (IC) with a defect-tolerant neural network. The neural network has one or more redundant neurons in some embodiments. After the IC is manufactured, a defective neuron in the neural network can be detected through a test procedure and then replaced by a redundant neuron (i.e., the redundant neuron can be assigned the operation of the defective neuron). The routing fabric of the neural network can be reconfigured so that it re-routes signals around the discarded, defective neuron. In some embodiments, the reconfigured routing fabric does not provide any signal to or forward any signal from the discarded, defective neuron, and instead provides signals to and forwards signals from the redundant neuron that takes the defective neuron's position in the neural network. In some embodiments that implement a neural network by re-purposing (i.e.

Abstract: An apparatus includes a processor to: train a first neural network of a chain to generate first configuration data including first trained parameters, wherein the chain performs an analytical function generating a set of output values from a set of input values, each neural network has inputs to receive the set of input values and outputs to output a portion of the set of output values, and the neural networks are ordered from the first at the head to a last neural network at the tail, and are interconnected so that each neural network additionally receives the outputs of a preceding neural network; train, using the first configuration data, a next neural network in the chain ordering to generate next configuration data including next trained parameters; and use at least the first and next configuration data and data indicating the interconnections to instantiate the chain to perform the analytical function.

Abstract: A method, system and computer-usable medium for providing cognitive insights comprising receiving streams of data from a plurality of data sources; processing the streams of data from the plurality of data sources, the processing the streams of data from the plurality of data sources performing data enriching to provide enriched data; generating the cognitive session graph, the cognitive session graph being associated with a session, the cognitive session graph comprising at least some enriched data; and, processing the cognitive session graph to provide a cognitive insight, the cognitive insight being related to the session.

Abstract: A neural network device having a first plurality of synapses that includes a plurality of memory cells. Each memory cell includes a floating gate over a first portion of a channel region and a first gate over a second portion of the channel region. The memory cells are arranged in rows and columns. A plurality of first lines each electrically connect together the first gates in one of the memory cell rows, a plurality of second lines each electrically connect together the source regions in one of the memory cell rows, and a plurality of third lines each electrically connect together the drain regions in one of the memory cell columns. The first plurality of synapses receives a first plurality of inputs as electrical voltages on the plurality of third lines, and provides a first plurality of outputs as electrical currents on the plurality of second lines.

Abstract: A system and method for pruning. A neural network includes a plurality of long short-term memory cells, each of which includes an input having a weight matrix Wc, an input gate having a weight matrix Wi, a forget gate having a weight matrix Wf, and an output gate having a weight matrix Wo. In some embodiments, after initial training, one or more of the weight matrices Wi, Wf, and Wo are pruned, and the weight matrix Wc is left unchanged. The neural network is then retrained, the pruned weights being constrained to remain zero during retraining.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for neural programming. One of the methods includes processing a current neural network input using a core recurrent neural network to generate a neural network output; determining, from the neural network output, whether or not to end a currently invoked program and to return to a calling program from the set of programs; determining, from the neural network output, a next program to be called; determining, from the neural network output, contents of arguments to the next program to be called; receiving a representation of a current state of the environment; and generating a next neural network input from an embedding for the next program to be called and the representation of the current state of the environment.

Abstract: Systems and methods for automatically self-correcting or correcting in real-time one or more neural networks after detecting a triggering event, or breaching boundary conditions are provided. Such a triggering event may indicate incorrect output signal or data being generated by the one or more neural networks. In particular, machine controllers of the invention limit the operations of neural networks to be within boundary conditions. Autonomous machines of the invention can be self-corrected after a breach of a boundary condition is detected. Autonomous land vehicles of the invention are capable of determining the timing of automatic transition to the manual control from automated driving mode. The controller of the invention filters and saves input-output data sets that fall within boundary conditions for later training of neural networks. The controllers of the invention include security architectures to prevent damages from virus attacks or system malfunctions.

Abstract: A radar system includes an array antenna including antenna elements that each output a reception signal in response to one or plural arriving waves, and a signal processing circuit in which a learned neural network has been established. The signal processing circuit receives the reception signals, inputs the reception signals or a secondary signal generated from the reception signals to the neural network, performs computation by using the reception signals or secondary signal and learned data of the neural network, and outputs a signal indicating the number of arriving waves from the neural network.

Abstract: A system includes a robot having a module that includes a function for mapping natural language commands of varying complexities to reward functions at different levels of abstraction within a hierarchical planning framework, the function including using a deep neural network language model that learns how to map the natural language commands to reward functions at an appropriate level of the hierarchical planning framework.

Abstract: A system for handling errors in a neural network includes a neural network processor for executing a neural network associated with use of a vehicle. The neural network processor includes an error detector configured to detect a data error associated with execution of the neural network and a neural network controller configured to receive a report of the data error from the error detector. In response to receiving the report, the neural network controller is further configured to signal that a pending result of the neural network is tainted without terminating execution of the neural network.

Abstract: Provided are a computer program product, a learning apparatus and a learning method. The method includes calculating, by a processor, a first propagation value that is propagated from a propagation source node to a propagation destination node in a neural network including a plurality of nodes, based on node values of the propagation source node at a plurality of time points and a weight corresponding to passage of time points based on a first attenuation coefficient. The method further includes updating, by the processor, a first update parameter, which is used for updating the first attenuation coefficient, by using the first propagation value. The method also includes updating, by the processor, the first attenuation coefficient by using the first update parameter and an error of the node value of the propagation destination node.

Abstract: A source system initializes, using an initialization seed, a first parameter vector representing weights of a neural network. The source system determines a second parameter vector by performing a sequence of mutations on the first parameter vector, the mutations each being based on a perturbation seed. The source system generates, and stores to memory, an encoded representation of the second parameter vector that comprises the initialization seed and a sequence of perturbation seeds corresponding to the sequence of mutations. The source system transmits the data structure to a target system, which processes a neural network based on the data structure.

Abstract: A method, computer program product, and system perform computations using a sparse convolutional neural network accelerator. A first vector comprising only non-zero weight values and first associated positions of the non-zero weight values within a 3D space is received. A second vector comprising only non-zero input activation values and second associated positions of the non-zero input activation values within a 2D space is received. The non-zero weight values are multiplied with the non-zero input activation values, within a multiplier array, to produce a third vector of products. The first associated positions are combined with the second associated positions to produce a fourth vector of positions, where each position in the fourth vector is associated with a respective product in the third vector. The products in the third vector are transmitted to adders in an accumulator array, based on the position associated with each one of the products.

Abstract: This document generally describes a neural network training system, including one or more computers, that trains a recurrent neural network (RNN) to receive an input, e.g., an input sequence, and to generate a sequence of outputs from the input sequence. In some implementations, training can include, for each position after an initial position in a training target sequence, selecting a preceding output of the RNN to provide as input to the RNN at the position, including determining whether to select as the preceding output (i) a true output in a preceding position in the output order or (ii) a value derived from an output of the RNN for the preceding position in an output order generated in accordance with current values of the parameters of the recurrent neural network.

Abstract: Systems and method are provided for controlling a vehicle. In one embodiment, a determination is made that a traffic control person and a traffic control sign are present within the environment of the vehicle based on sensor data, such as optical camera data. The position and orientation of the traffic control sign relative to the traffic control person is determined, e.g., via lidar sensor data, and the validity of the traffic control person and the traffic control sign is determined based on the position and orientation of the traffic control sign.

Abstract: Systems and methods for an interconnection scheme for reconfigurable neuromorphic hardware are disclosed. A neuromorphic processor may include a plurality of corelets, each corelet may include a plurality of synapse arrays and a neuron array. Each synapse array may include a plurality of synapses and a synapse array router coupled to synapse outputs in a synapse array. Each synapse may include a synapse input, synapse output; and a synapse memory. A neuron array may include a plurality of neurons, each neuron may include a neuron input and a neuron output. Each synapse array router may include a first logic to route one or more of the synapse outputs to one or more of the neuron inputs.

Abstract: A method of conditioning one or more actions related to an operator controllable device is disclosed. The method includes receiving one or more action-based instructions representing one or more operator controllable device actions associated with the operator controllable device, and receiving one or more action-oriented time representations representing one or more times associated with the action-based instructions. The method also includes deriving at least one conditioned action-based instruction for a time subsequent to the times represented by the received action-oriented time representations from: the received action-based instructions, and the received action-oriented time representations. The conditioned action-based instruction, when executed, causes the operator controllable device to take one or more conditioned actions.

Abstract: A neural network unit includes first and second memories that hold rows of respective N weight and data words and provides a row of them to N corresponding neural processing units (NPU), respectively. The N NPUs each have an accumulator and an arithmetic unit that performs a series of multiply operations on pairs of weight words and data words received from the first and second memories to generate a series of products. The arithmetic unit also performs a series of addition operations on the series of products to accumulate an accumulated value in the accumulator. Activation function units (AFU) are each shared by a corresponding plurality of the N NPUs. Each AFU, in a sequential fashion with respect to each NPU of the corresponding plurality of the N NPUs, receives the accumulated value from the NPU and performs an activation function on the accumulated value to generate a result.

Abstract: Methods, apparatus, systems, and computer program products are provided for providing distributed online learning for personal predictive models that preserves user privacy. An example method comprises receiving model updates based at least in part on synchronization metadata. The model update comprises one or more received model parameter updates. The received model parameter updates correspond to one or more model parameters. The model updates are associated with a particular apparatus. The example method further comprises determining a decay factor based at least in part on the synchronization metadata; applying the decay factor to the one or more received model parameter updates; and updating the one or more model parameters based on the decayed model parameter updates.

Abstract: An apparatus includes: an adaptive declustered RAID array configured of data storage devices (DSDs), the DSDs comprise data chunks allocated as data, a local parity, or a global parity; and a controller configured to generate a reliability indicator reflective of a reliability status of at least a portion of the adaptive declustered RAID array for reallocating the data chunks by dynamically increasing or decreasing the data chunks allocated as the local parity, the global parity, or a combination thereof.

Abstract: A device configured to emulate a correlithm object logic function gate comprises a memory and a logic engine. The memory stores a logical operator truth table that includes a plurality of input logical values, a plurality of output logical values, and a plurality of logical operators. These logical values and the logical operators are represented by correlithm objects. The logic engine receives at least one input and a context input correlithm object representing one of the plurality of logical operators. The logic engine determines a portion of the truth table to apply based at least in part upon the logical operator represented by the context input correlithm object. The logic engine further determine an output of the logic function gate based at least in part upon the determined portion of the truth table to apply and the input.

Abstract: A calculation device includes an adding unit configured to add at least one new node to a network, which has multiple nodes that output results of calculations on input data are connected and which learned a feature of data belonging to a first subclass contained in a predetermined class. The calculation device includes an accepting unit configured to accept, as input data, training data belonging to a second subclass contained in the predetermined class. The calculation device includes a calculation unit configured to calculate coupling coefficients between the new node added by the adding unit and other nodes to learn a feature of the training data belonging to the second subclass based on an output result obtained when the training data accepted by the accepting unit is input to the network.

Abstract: The Developmental Network incorporates a Turing Machine that injects teaching instructions directly into the skull-closed network. The Developmental Network can also autonomously learn directly from the natural world without the need for a human to encode its input and output. The neural network so configured can be used as a controller for robotic and other computer control applications where the neural network is organized into plural X-Y-Z areas receiving signals from sensors and providing signals to effectors.

Abstract: An artificial intelligence inference computing device contains a printed circuit board (PCB) and a number of electronic components mounted thereon. Electronic components include a wireless communication module, a controller module, a memory module, a storage module and at least one cellular neural networks (CNN) based integrated circuit (IC) configured for performing convolutional operations in a deep learning model for extracting features out of input data. Each CNN based IC includes a number of CNN processing engines operatively coupled to at least one input/output data bus. CNN processing engines are connected in a loop with a clock-skew circuit. Wireless communication module is configured for transmitting pre-trained filter coefficients of the deep learning model, input data and classification results.

Abstract: The disclosed subject matter involves identifying clusters and segments of a population of data for use in a recommendation service. Clusters of members or items are formed, where the clusters, or partitions are close to being equal in size. Items are distributed based on similarities identified with matrix factorization. The items are formed into clusters based on the similarities and the clusters are used in training of a generalized linear mixed model treating the clusters as random-level effects. The trained model may be used in the recommendation service. Other embodiments are described and claimed.

Abstract: A method for identifying frequently occurring waveform patterns in time series comprises segmenting each of one or more time series into a plurality of subsequences. Further, a subsequence matrix comprising each of the plurality of subsequences is generated. Further, the subsequence matrix is processed to obtain a candidate subsequence matrix comprising a plurality of non-trivial subsequences. Further, the plurality of non-trivial subsequences is clustered into a plurality of spherical clusters of a predetermined diameter. Further, a plurality of sub-clusters for each of one or more spherical clusters is obtained based on a mean of each of the plurality of non-trivial subsequences present in the spherical cluster. Further, one or more frequent waveform clusters, depicting frequently occurring waveform patterns, are ascertained from amongst the one or more spherical clusters based on a number of non-trivial subsequences present in each of the plurality of sub-clusters of the spherical cluster.

Abstract: A neuromorphic synapse with a resistive memory cell connected in circuitry having first and second input terminals. The input terminals respectively receive pre-neuron and post-neuron action signals, each having a read portion and a write portion, in use. The circuitry includes an output terminal for providing a synaptic output signal which is dependent on resistance of the memory cell. The circuitry is configured such that the synaptic output signal is provided at the output terminal in response to application at the first input terminal of the read portion of the pre-neuron action signal, and such that a programming signal, for programming resistance of the memory cell, is applied to the cell in response to simultaneous application of the write portions of the pre-neuron and post-neuron action signals at the first and second input terminals respectively. The synapse can be adapted for operation with identical pre-neuron and post-neuron action signals.

Abstract: The disclosed subject matter involves identifying clusters and segments of a population of data for use in a recommendation service. Clusters of members or items are formed, where the clusters, or partitions are close to being equal in size, items are distributed based on similarities identified with matrix factorization. A matrix used in the matrix factorization is customized based on the recommendation type. The items are formed into clusters based on the similarities and the clusters are used in training of a generalized linear mixed model treating the clusters as random-level effects. The trained model may be used in the recommendation service. Other embodiments are described and claimed.

Abstract: Systems, components, and methodologies are provided for improvements in operation of automotive vehicles by enabling emulation of traffic signal operation by genertic algorithms, providing tunable solutions for efficient and safe operation.

Abstract: A convolution engine, such as a convolution neural network, operates efficiently with respect to sparse kernels by implementing zero skipping. An input tile is loaded and accumulated sums are calculated for the input tile for non-zero coefficients by shifting the tile according to a row and column index of the coefficient in the kernel. Each coefficient is applied individually to tile and the result written to an accumulation buffer before moving to the next non-zero coefficient. A 3D or 4D convolution may be implemented in this manner with separate regions of the accumulation buffer storing accumulated sums for different indexes along one dimension. Images are completely processed and results for each image are stored in the accumulation buffer before moving to the next image.

Abstract: Sparse representation of information performs powerful feature extraction on high-dimensional data and is of interest for applications in signal processing, machine vision, object recognition, and neurobiology. Sparse coding is a mechanism by which biological neural systems can efficiently process complex sensory data while consuming very little power. Sparse coding algorithms in a bio-inspired approach can be implemented in a crossbar array of memristors (resistive memory devices). This network enables efficient implementation of pattern matching and lateral neuron inhibition, allowing input data to be sparsely encoded using neuron activities and stored dictionary elements. The reconstructed input can be obtained by performing a backward pass through the same crossbar matrix using the neuron activity vector as input. Different dictionary sets can be trained and stored in the same system, depending on the nature of the input signals.

Abstract: The disclosed embodiments provide a system for processing data. During operation, the system obtains a global version of a statistical model and a user-specific version of the statistical model for a user. Next, the system applies the global version to member features of the user and job features of a set of jobs to generate a first ranking of the jobs for the user. The system then applies the user-specific version to the member features and the job features for a highest-ranked subset of jobs in the first ranking to generate a second ranking of the jobs for the user. Finally, the system outputs at least a portion of the second ranking as a set of job recommendations.

Abstract: Embodiments of the invention provide a method for mapping a bipartite graph onto a neuromorphic architecture comprising of a plurality of interconnected neuromorphic core circuits. The graph includes a set of source nodes and a set of target nodes. The method comprises, for each source node, creating a corresponding splitter construct configured to duplicate input. Each splitter construct comprises a first portion of a core circuit. The method further comprises, for each target node, creating a corresponding merger construct configured to combine input. Each merger construct comprises a second portion of a core circuit. Source nodes and target nodes are connected based on a permutation of an interconnect network interconnecting the core circuits.