Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a neural network used to select actions performed by a reinforcement learning agent interacting with an environment. In one aspect, a method includes maintaining a replay memory, where the replay memory stores pieces of experience data generated as a result of the reinforcement learning agent interacting with the environment. Each piece of experience data is associated with a respective expected learning progress measure that is a measure of an expected amount of progress made in the training of the neural network if the neural network is trained on the piece of experience data. The method further includes selecting a piece of experience data from the replay memory by prioritizing for selection pieces of experience data having relatively higher expected learning progress measures and training the neural network on the selected piece of experience data.

Abstract: A machine learning management apparatus identifies a maximum prediction performance score amongst a plurality of prediction performance scores corresponding to a plurality of models generated by executing each of a plurality of machine learning algorithms. As for a first machine learning algorithm having generated a model corresponding to the maximum prediction performance score, the machine learning management apparatus determines a first training dataset size to be used when the first machine learning algorithm is executed next time based on the maximum prediction performance score, first estimated prediction performance scores, and first estimated runtimes.

Abstract: A computing system including one or more processors configured to receive an audio input. The one or more processors may generate a text transcription of the audio input at a sequence-to-sequence speech recognition model, which may assign a respective plurality of external-model text tokens to a plurality of frames included in the audio input. Each external-model text token may have an external-model alignment within the audio input. Based on the audio input, the one or more processors may generate a plurality of hidden states. Based on the plurality of hidden states, the one or more processors may generate a plurality of output text tokens. Each output text token may have a corresponding output alignment within the audio input. For each output text token, a latency between the output alignment and the external-model alignment may be below a predetermined latency threshold. The one or more processors may output the text transcription.

Abstract: In a method of training a DNN, a weight matrix (W) is provided as a linear combination of matrices/arrays A and C. In a forward cycle, an input vector x is transmitted through arrays A and C and output vector y is read. In a backward cycle, an error signal ? is transmitted through arrays A and C and output vector z is read. Array A is updated by transmitting input vector x and error signal ? through array A. In a forward cycle, an input vector ei is transmitted through array A and output vector y? is read. ƒ(y?) is calculated using y?. Array C is updated by transmitting input vector ei and ƒ(y?) through array C. A DNN is also provided.

Abstract: Described herein are systems and methods for compressing and speeding up dense matrix multiplications as found, for examples, in the fully connected and recurrent layers of neural networks for embedded large vocabulary continuous speech recognition (LVCSR). For compression, trace norm regularization technique embodiments were introduced and studied for training low rank factored versions of matrix multiplications. Compared to standard low rank training, the methods more consistently lead to good accuracy versus number of parameter trade-offs and can be used to speed-up training of large models. Faster inference may be further enabled on ARM processors through kernels optimized for small batch sizes, resulting in speed ups over the currently used library. Beyond LVCSR, the techniques are also generally applicable to embedded neural networks with large fully connected or recurrent layers.

Abstract: The embodiments herein describe hybrid parallelism techniques where a mix of data and model parallelism techniques are used to split the workload of a layer across an array of processors. When configuring the array, the bandwidth of the processors in one direction may be greater than the bandwidth in the other direction. Each layer is characterized according to whether they are more feature heavy or weight heavy. Depending on this characterization, the workload of an NN layer can be assigned to the array using a hybrid parallelism technique rather than using solely the data parallelism technique or solely the model parallelism technique. For example, if an NN layer is more weight heavy than feature heavy, data parallelism is used in the direction with the greater bandwidth (to minimize the negative impact of weight reduction) while model parallelism is used in the direction with the smaller bandwidth.

Abstract: A method and system of training a natural language processing network are provided. A corpus of data is received and one or more input features selected therefrom by a generator network. The one or more selected input features from the generator network are received by a first predictor network and used to predict a first output label. A complement of the selected input features from the generator network are received by a second predictor network and used to predict a second output label.

Abstract: A parallel recursive neural network, including: a memory configured to store data and processing instructions; and a parallel computer processor configured to: receive a set of input values; apply a recursive layer function individually on each of the set of input values in parallel to produce a set of hidden states; apply a reduction function on pairs of adjacent hidden states in the set of hidden states in parallel to produce a new set of hidden states; and repeat applying the reduction function of pairs of adjacent states in the new set of hidden states in parallel until a single output hidden state results.

Abstract: This disclosure relates to method and system for improving performance of an artificial neural network (ANN). The method may include generating a weight matrix comprising weights of neural nodes in a given layer for each layer of the ANN, determining a marginal contribution value of a given neural node for each neural node in the given layer with respect to other neural nodes in the given layer, executing an elimination decision for each neural node in each layer based on the corresponding marginal contribution value, determining a distributed surplus value of a given remaining neural node in a given layer based on the marginal contribution values of a coalition of remaining neural nodes in the given layer for each remaining neural node in each layer, and updating the weight matrix based on the distributed surplus value of each remaining neural node in each layer.

Abstract: A method for optimizing a data model is used in a device. The device acquires data information and selecting at least two data models according to the data information, and utilizes the data information to train the at least two data models. The device acquires each accuracy of the at least two data models, determines a target data model which has greatest accuracy between the at least two data models, and optimizes the target data model.

Abstract: A method of continual learning in an artificial intelligence system through bi-level optimization includes providing a stored data sample of a current task and providing a neural network subdivided into two parts including a parameter part and a hyper-parameter part. The method further includes performing bi-level optimization by separately training the two parts of the neural network. The neural network has been trained, prior to the bi-level optimization, on data samples of previous tasks.

Abstract: Provided are a method and apparatus for constructing a compact translation model that may be installed on a terminal on the basis of a pre-built reference model, in which a pre-built reference model is miniaturized through a parameter imitation learning and is efficiently compressed through a tree search structure imitation learning without degrading the translation performance. The compact translation model provides translation accuracy and speed in a terminal environment that is limited in network, memory, and computation performance.

Abstract: Training a generator G of a GAN includes generating, by G and in response to receiving a first input Z, a first output G(Z); generating, by an encoder E of the GAN and in response to receiving the first output G(Z) as input, a second output E(G(Z)); generating, by G and in response to receiving the second output E(G(Z)) as input, a third output G(E(G(Z))); generating, by E and in response to receiving the third output G(E(G(Z))) as input, a fourth output E(G(E(G(Z)))); training E to minimize a difference between the second output E(G(Z)) and the fourth output E(G(E(G(Z)))); and using the second output E(G(Z)) and fourth output E(G(E(G(Z)))) to constrain a training of the generator G. G(Z) is an ambient space representation Z. E(G(Z)) is a latent space representation of G(Z). G(E(G(Z))) is an ambient space representation of E(G(Z)). E(G(E(G(Z)))) is a latent space representation of G(E(G(Z))).

Abstract: Methods and systems for solving a linear system include setting resistances in an array of settable electrical resistances in accordance with values of an input matrix. A series of input vectors is applied to the array as voltages to generate a series of respective output vectors. Each input vector in the series of vectors is updated based on comparison of the respective output vectors to a target vector. A solution of a linear system is determined that includes the input matrix based on the updated input vectors.

Abstract: Systems and methods for training a multitask network is disclosed. In one aspect, training the multitask network includes determining a gradient norm of a single-task loss adjusted by a task weight for each task, with respect to network weights of the multitask network, and a relative training rate for the task based on the single-task loss for the task. Subsequently, a gradient loss function, comprising a difference between (1) the determined gradient norm for each task and (2) a corresponding target gradient norm, can be determined. An updated task weight for the task can be determined and used in the next iteration of training the multitask network, using a gradient of the gradient loss function with respect to the task weight for the task.

Abstract: There are provided a neural network weight discretizing method, system and device, and a computer readable storage medium. The method includes acquiring a weight value range and a number of discrete weight states, the weight value range referring to a range of discrete weight values consisting of a maximum weight value of a current time step and a minimum weight value of the current time step, and the number of discrete weight states referring to the quantity of discrete weight states. The method also includes acquiring a weight state of a previous time step and a weight increment of the current time step and acquiring a state transfer direction by using a directional function according to the weight increment of the current time step. The method also includes acquiring a weight state of the current time step according to the weight state of the previous time step, the weight increment of the current time step, the state transfer direction, the weight value range and the number of discrete weight states.

Abstract: Methods, apparatus, systems and articles of manufacture for federated training of a neural network using trusted edge devices are disclosed. An example system includes an aggregator device to aggregate model updates provided by one or more edge devices. The one or more edge devices to implement respective neural networks, and provide the model updates to the aggregator device. At least one of the edge devices to implement the neural network within a trusted execution environment.

Abstract: Described is a system for specifying control of a device based on a Bayesian network model. The system includes a Bayesian neuromorphic compiler having a network composition module having probabilistic computation units (PCUs) arranged in a hierarchical composition containing multi-level dependencies. The Bayesian neuromorphic compiler receives a Bayesian network model as input and produces a spiking neural network topology and configuration that implements the Bayesian network model. The network composition module learns conditional probabilities of the Bayesian network model. The system computes a conditional probability and controls a device based on the computed conditional probability.

Abstract: According to various embodiments, a method for generating one or more optimal neural network architectures is disclosed. The method includes providing an initial seed neural network architecture and utilizing sequential phases to synthesize the neural network until a desired neural network architecture is reached. The phases include a gradient-based growth phase and a magnitude-based pruning phase.

Abstract: A learning agent is disclosed that receives data in sequence from one or more sequential data sources; generates a model modelling sequences of data and actions; and selects an action maximizing the expected future value of a reward function, wherein the reward function depends at least partly on at least one of: a measure of the change in complexity of the model, or a measure of the complexity of the change in the model. The measure of the change in complexity of the model may be based on, for example, the change in description length of the first part of a two-part code describing one or more sequences of received data and actions, the change in description length of a statistical distribution modelling, the description length of the change in the first part of the two-part code, or the description length of the change in the statistical distribution modelling.

Abstract: Predetermined information for determining whether an input screen on which a user inputs an evaluation of a program is to be displayed is received from a server. If the received predetermined information indicates that the input screen is to be displayed, the input screen is displayed on a display included in an information processing apparatus.

Abstract: Provided is an artificial intelligence (AI) apparatus for managing an operation of an artificial intelligence (AI) system. The AI apparatus includes: a communication unit that receives state information from at least one member AI apparatus included in the AI system, respectively; a memory that stores apparatus information on the at least one member AI apparatus, respectively; and a processor that: upon acquiring a control command of a user, determines a target member AI apparatus to perform the control command; determines whether the target member AI apparatus is capable of performing the control command or not; transmits the control command to the target member AI apparatus if the target member AI apparatus is capable of performing the control command, and outputs a response indicating that the target member AI apparatus is not capable of performing the control command if the target member AI apparatus is not capable of performing the control command.

Abstract: The present specification discloses a trusted execution environment (TEE)-based model training method and apparatus. In one or more embodiments, the method includes: obtaining encrypted target samples from an encrypted training sample set in a first execution environment, inputting the encrypted target samples into a second execution environment that is a trusted execution environment (TEE) different from the first execution environment, decrypting the encrypted target samples in the TEE to obtain decrypted target samples, inputting the decrypted target samples into a feature extraction model in the TEE to determine sample features, determining the sample features output from the TEE as target sample features for a current iteration of a training process for a target model, and performing, based on the target sample features, the current iteration on the target model in the first execution environment.

Abstract: A memory-centric neural network system and operating method thereof includes: a processing unit; semiconductor memory devices coupled to the processing unit, the semiconductor memory devices containing instructions executed by the processing unit; a weight matrix constructed with rows and columns of memory cells, inputs of the memory cells of a same row being connected to one of axons, outputs of the memory cells of a same column being connected to one of neurons; timestamp registers registering timestamps of the axons and the neurons; and a lookup table containing adjusting values indexed in accordance with the timestamps, wherein the processing unit updates the weight matrix in accordance with the adjusting values.

Abstract: Systems and methods analyze training of a first machine learning system with a second machine learning system. The first machine learning system comprises a neural network with a first inner layer node. The method includes connecting the first machine learning system to an input of the second machine learning system. The second machine learning system comprises a second objective function for analyzing an internal characteristic of the first machine learning system and which is different from a first objective function for the first machine learning system.

Abstract: A memory-centric neural network system and operating method thereof includes: a processing unit; semiconductor memory devices coupled to the processing unit, the semiconductor memory devices containing instructions executed by the processing unit; a weight matrix constructed with rows and columns of memory cells, inputs of the memory cells of a same row being connected to one of axons, outputs of the memory cells of a same column being connected to one of neurons; timestamp registers registering timestamps of the axons and the neurons; and a lookup table containing adjusting values indexed in accordance with the timestamps, wherein the processing unit updates the weight matrix in accordance with the adjusting values.

Abstract: Disclosed is a convolution operator system for performing a convolution operation concurrently on an image. An input router receives image data. A controller allocates image data to a set of computing blocks based on the size of the image data and number of available computing blocks. Each computing block produces a convolution output corresponding to each row of the image. The controller allocates a plurality of group having one or more computing blocks to generate a set of convolution output. Further, a pipeline adder aggregates the set of convolution output to produce an aggregated convolution output. An output router transmits either the convolution output or the aggregated convolution output for performing subsequent convolution operation to generate a convolution result for the image data.

Abstract: Method or system for reinforcement learning that simultaneously learns a DR distribution ? while optimizing an agent policy ? to maximize performance over the learned DR distribution; method or system for training a learning agent using data synthesized by a simulator based on both a performance of the learning agent and a range of parameters present in the synthesized data.

Abstract: A control system for automatic operation of a coker, the control system. The control system includes a drum feeder operable to modulate a feed of oil into a coke drum of the coker. The control system further includes a controller with a processing circuit. The processing circuit obtains a target coke rate indicating a target rate at which to accumulate coke within the coke drum. The processing circuit further uses a neural network model to generate a target coker feed rate predicted to result in the coke accumulating within the coke drum at the target coke rate. The target coker feed rate indicates a target rate at which to feed the oil into the coke drum. The processing circuit further operates the drum feeder using the target coker feed rate to modulate the feed of oil into the coke drum.

Abstract: Methods and systems are provided for generating a multi-label classification system. The multi-label classification system can use a multi-label classification neural network system to identify one or more labels for an image. The multi-label classification system can explicitly take into account the relationship between classes in identifying labels. A relevance sub-network of the multi-label classification neural network system can capture relevance information between the classes. Such a relevance sub-network can decouple independence between classes to focus learning on relevance between the classes.

Abstract: A system, method and non-transitory computer readable medium for editing images with verbal commands are described. Embodiments of the system, method and non-transitory computer readable medium may include an artificial neural network (ANN) comprising a word embedding component configured to convert text input into a set of word vectors, a feature encoder configured to create a combined feature vector for the text input based on the word vectors, a scoring layer configured to compute labeling scores based on the combined feature vectors, wherein the feature encoder, the scoring layer, or both are trained using multi-task learning with a loss function including a first loss value and an additional loss value based on mutual information, context-based prediction, or sentence-based prediction, and a command component configured to identify a set of image editing word labels based on the labeling scores.

Abstract: Systems and methods are described for matching a corrupted database record with a record of a validated database. The system receives a corrupted record from a first database. The corrupted record is vectorized to create an input data vector. A denoised data vector is generated by applying a denoising autoencoder to the input data vector, where the denoising autoencoder is specific to the first database. The system compares the denoised data vector with each of a plurality of validated data vectors generated based on records of the validated database to determine that a first denoised data vector matches a matching vector. In response, the system trains the denoising autoencoder using a data pair that includes the input data vector and the matching vector. The system also outputs the validated record that was used to generate the first matching vector.

Abstract: A novel architecture for a network of deep reinforcement modules that enables cross-functional and multi-system coordination of autonomous systems for self-optimization with a reduced computational footprint is disclosed. Each deep reinforcement module in the network is comprised of either a single artificial neural network or a deep reinforcement module sub-network. DReMs are designed independently, decoupling each requisite function. Each module of a deep reinforcement module network is trained independently through deep reinforcement learning. By separating the functions into deep reinforcement modules, reward functions can be designed for each individual function, further simplifying the development of a full suite of algorithms while also minimizing training time.

Abstract: A method of pruning a batch normalization layer from a pre-trained deep neural network model is proposed. The pre-trained deep neural network model is inputted as a candidate model. The candidate model is pruned by removing the at least one batch normalization layer from the candidate model to form a pruned candidate model only when the at least one batch normalization layer is connected to and adjacent to a corresponding linear operation layer. The corresponding linear operation layer may be at least one of a convolution layer, a dense layer, a depthwise convolution layer, and a group convolution layer. Weights of the corresponding linear operation layer are adjusted to compensate for the removal of the at least one batch normalization. The pruned candidate model is then output and utilized for inference.

Abstract: A machining condition adjustment device includes a data acquisition unit that acquires at least one piece of data indicating a state of machining including a machining type in a machine tool, a priority condition storage unit that stores priority condition data in which the machining type is associated with a priority condition, a preprocessing unit that produces data to be used for machine learning, and a machine learning device that carries out processing of the machine learning related to at least either of a machining condition and a machining parameter for machining by the machine tool. The machine learning device includes a learning model storage unit that stores a plurality of learning models generated for each machining type and a learning model selection unit that selects a learning model, based on the machining type.

Abstract: Embodiments of the invention are directed to systems, methods, and computer program products for an electronic system for management of image processing model database. The system is configured for versioning machine-learning neural-network based image processing models and identifying and tracking mutations in hyper parameters amongst versions of image processing models. The system is configured to determine that a second image processing model is a version of a first image processing model. The system is further configured to map the mutations in hyper parameters between the first plurality of hyper parameters of the first image processing model and the second plurality of hyper parameters associated with the second image processing model.

Abstract: Various examples described herein are directed to systems and methods that learn to evaluate data items. A computing device receives data items and evaluates the data items. The evaluation performed by the computing device includes comparing a first data item with a second data item and determining a difference between the first data item and the second data item based on the comparison. The computing device also prepares a recommendation based on the difference between the first data item and the second data item and forwards the recommendation to a subject matter expert. The computing device receives an updated recommendation from the subject matter expert that is based on the recommendation. Using the updated recommendation, the computing device refines the comparing and determining operations where the computing device further learns to evaluate the data items by receiving the updated recommendation and refining the comparing and determining operations.

Abstract: A system for training an artificial intelligence (AI) model for an AI chip to implement an AI task may include an AI training unit to train weights of an AI model in floating point, a convolution quantization unit for quantizing the trained weights to a number of quantization levels, and an activation quantization unit for updating the weights of the AI model so that output of the AI model based at least on the updated weights are within a range of activation layers of the AI chip. The updated weights may be stored in fixed point and uploadable to the AI chip. The various units may be configured to account for the hardware constraints in the AI chip to minimize performance degradation when the trained weights are uploaded to the AI chip and expedite training convergence. Forward propagation and backward propagation may be combined in training the AI model.

Abstract: An off-policy reinforcement learning actor-critic neural network system configured to select actions from a continuous action space to be performed by an agent interacting with an environment to perform a task. An observation defines environment state data and reward data. The system has an actor neural network which learns a policy function mapping the state data to action data. A critic neural network learns an action-value (Q) function. A replay buffer stores tuples of the state data, the action data, the reward data and new state data. The replay buffer also includes demonstration transition data comprising a set of the tuples from a demonstration of the task within the environment. The neural network system is configured to train the actor neural network and the critic neural network off-policy using stored tuples from the replay buffer comprising tuples both from operation of the system and from the demonstration transition data.

Abstract: A micro-processor circuit and a method of performing neural network operation are provided. The micro-processor circuit is suitable for performing neural network operation. The micro-processor circuit includes a parameter generation module, a compute module and a compare logic. The parameter generation module receives in parallel a plurality of input parameters and a plurality of weight parameters of the neural network operation. The parameter generation module generates in parallel a plurality of sub-output parameters according to the input parameters and the weight parameters. The compute module receives in parallel the sub-output parameters. The compute module sums the sub-output parameters to generate a summed parameter. The compare logic receives the summed parameter. The compare logic performs a comparison operation based on the summed parameter to generate a plurality of output parameters of the neural network operation.

Abstract: Disclosed herein are system, method, and computer program product embodiments for an improved spiking neural network (SNN) configured to learn and perform unsupervised extraction of features from an input stream. An embodiment operates by receiving a set of spike bits corresponding to a set synapses associated with a spiking neuron circuit. The embodiment applies a first logical AND function to a first spike bit in the set of spike bits and a first synaptic weight of a first synapse in the set of synapses. The embodiment increments a membrane potential value associated with the spiking neuron circuit based on the applying. The embodiment determines that the membrane potential value associated with the spiking neuron circuit reached a learning threshold value. The embodiment then performs a Spike Time Dependent Plasticity (STDP) learning function based on the determination that the membrane potential value of the spiking neuron circuit reached the learning threshold value.

Abstract: Provided are a storage device using a neural network and an operating method of the storage device for automatic redistribution of information and variable storage capacity based on accuracy-storage capacity tradeoff that may learn input information using the neural network and may store the learned information. The neural network may include a plurality of input neurons and a plurality of output neurons; at least one stable synapse configured to connect at least one of the input neurons and at least one of the output neurons, respectively; and at least one flexible synapse configured to connect at least one remaining of the input neurons and at least one remaining of the output neurons, respectively.

Abstract: Methods are provided for implementing training of a machine learning model in a processing system, together with systems for performing such methods. A method includes providing a core module for effecting a generic optimization process in the processing system, and in response to a selective input, defining a set of derivative modules, for effecting computation of first and second derivatives of selected functions ƒ and g in the processing system, to be used with the core module in the training operation. The method further comprises performing, in the processing system, the generic optimization process effected by the core module using derivative computations effected by the derivative modules.

Abstract: A memory network can be constructed with at least memory write weightings, memory read weightings and at least one read vector, the memory write weightings parameterizing memory write operations of a neural network to the memory matrix, the memory read weightings parameterizing memory read operations of the neural network from the memory matrix. At least one of the write weightings, the read weightings, or elements of the at least one read vector, can be initialized to have sparsity and/or low discrepancy sampling pattern. The memory network can be trained to perform a task.

Abstract: A method includes: providing a deep neural networks (DNN) model comprising a plurality of layers, each layer of the plurality of layers includes a plurality of nodes; sampling a change of a weight for each of a plurality of weights based on a distribution function, each weight of the plurality of weights corresponds to each node of the plurality of nodes; updating the weight with the change of the weight multiplied by a sign of the weight; and training the DNN model by iterating the steps of sampling the change and updating the weight. The plurality of weights has a high rate of sparsity after the training.

Abstract: A method of controlling an information processing apparatus, the information processing apparatus being configured to perform learning processing by using a neural network, the method includes: executing a calculation processing that includes calculating a learning rate, the learning rate being configured to change in the form of a continuous curve such that the time from when the learning rate is at an intermediate value of a maximum value to when the learning rate reaches a minimum value is shorter than the time from when the learning processing starts to when the learning rate reaches the intermediate value of the maximum value; and executing a control processing that includes controlling, based on the calculated learning rate, an amount of update at the time when a weight parameter is updated in an update processing.

Abstract: A computer-implemented method, medium, and system are disclosed. One example method includes determining multiple model bases by multiple service parties. A respective local service model is constructed by each service party. Respective local training samples are processed by each service party using the respective local service model to determine respective gradient data corresponding to each model basis. The respective gradient data is sent to a server. In response to determining that the first model basis satisfies a gradient update condition, corresponding gradient data of the first model basis received from each service party are combined to obtain global gradient data corresponding to the first model basis. The global gradient data is sent to each service party. Reference parameters in local model basis corresponding to the first model basis are updated by each service party using the global gradient data to train the respective local service model.

Abstract: A method comprising: obtaining a block of a picture or a picture in an encoder; determining if the block/picture is used for on-line learning; if affirmative, encoding the block/picture; reconstructing a coarse version of the block/picture or the respective prediction error block/picture; enhancing the coarse version using a neural net; fine-tuning the neural net with a training signal based on the coarse version; determining if the block/picture is enhanced using the neural net; and if affirmative, encoding the block/picture with enhancing using the neural net.

Abstract: An artificial intelligence platform system includes at least a server designed and configured to receive training data. Receiving training data includes receiving a first training set including a plurality of first data entries, each first data entry of the plurality of first data entries including at least an element of user data and at least a correlated first constitutional label. At least a server receives at least a user input datum from a user client device. At least a server generates at least an output as a function of the at least a user input datum and the training data. At least a server retrieves at least a stored user datum as a function of the at least a user input datum and the at least an output. At least a server transmits the at least a stored user datum to a user client device.

Abstract: This disclosure relates to method and system for verifying a positive classification performed by an artificial neural network (ANN) in a given class. The method includes generating a weight matrix comprising a weight of each neural node in a given layer; determining a contribution factor of a given neural node in the given layer with respect to an output of the ANN for the given class based on a known input vector to the given layer and a modified weight matrix; and generating a dominance matrix based on the contribution factor of each neural node in the given layer. The method further includes determining a rank of each neural node based on the corresponding dominance factor; and verifying the positive classification performed by the ANN in the given class for a test input vector based on the rank of each neural node in each layer of the ANN.