Abstract: A system and method for generating approximations of query results. The method includes sending a received query to a neural network, wherein the received query is executable on a target data set; receiving from the neural network a predicted result to the received query; providing the predicted result as a first output to a device having initiated the received query; determining a real result of the query from a data set stored in the database when the predicted result is insufficiently accurate; and providing the real result as a second output to a device having initiated the received query.

Abstract: An image processing apparatus obtains a first output image by applying an image to a first training network model, obtains a second output image by applying the image to a second training network model, and obtains a reconstructed image based on the first output image and the second output image. The first training network model is a model that uses a fixed parameter obtained through training of a plurality of sample images, the second training network model is trained to minimize a difference between a target image corresponding to the image and the reconstructed image.

Abstract: Provided are an application prediction method, an application preloading method and an application preloading apparatus. The application prediction method includes: obtaining a user behavior sample in a preset time period, where the user behavior sample includes an association record of usage timing of at least two applications determined from two or more applications on a terminal including a usage record of the at least two application and a usage timing relationship of the at least two applications; and training a preset prediction model according to the association record of usage timing to generate an application prediction model, thereby may take full advantage of the association record of usage timing of the applications which may truly reflect the user behavior, optimize the application preloading mechanism, improve the accuracy of the prediction of the application to be started effectively, and further reduce power consumption of the terminal system and the memory usage.

Abstract: A method may include receiving a graph-based model in a first format, including a static topology of the graph-based model. The method may also include encoding the graph-based model from the first format into a neural network topology optimizer (NNTO) readable format such that the topology of the encoded graph-based model is configured to be altered; creating a first group of entities based on at least a same portion of the encoded graph-based model; and performing a learning operation by tuning parameters of the first group of entities to produce an optimization score for each entity. Additionally, the method may include performing a validation operation; determining that an improvement in validation performance for at least one entity is within a threshold amount of improvement; selecting a solution entity; and adding the selected solution entity into the graph-based model in place of the same portion.

Abstract: Exemplary embodiments relate to adapting a generic language model during runtime using domain-specific language model data. The system performs an audio frame-level analysis, to determine if the utterance corresponds to a particular domain and whether the ASR hypothesis needs to be rescored. The system processes, using a trained classifier, the ASR hypothesis (a partial hypothesis) generated for the audio data processed so far. The system determines whether to rescore the hypothesis after every few audio frames (representing a word in the utterance) are processed by the speech recognition system.

Abstract: Systems and methods for forming radio frequency beams in communication systems are provided. Signals from one or more devices are received at a base station and are processed using a vector based deep learning (VBDL) model or network. The VBDL model can receive and process vector and/or spatial information related to or part of the received signals. An optimal beamforming vector for a received signal is determined by the VBDL model, without reference to a codebook. The VBDL model can incorporate parameters that are pruned during training to provide efficient operation of the model.

Abstract: Embodiments of the present invention relate to providing fault-tolerant power minimization in a multi-core neurosynaptic network. In one embodiment of the present invention, a method of and computer program product for fault-tolerant power-driven synthesis is provided. Power consumption of a neurosynaptic network is modeled as wire length. The neurosynaptic network comprises a plurality of neurosynaptic cores connected by a plurality of routers. At least one faulty core of the plurality of neurosynaptic cores is located. A placement blockage is modeled at the location of the at least one faulty core. A placement of the neurosynaptic cores is determined by minimizing the wire length.

Abstract: A method for calculating an output of a neural network, including the steps of generating a first neural network that includes discrete edge weights from a neural network that includes precise edge weights by stochastic rounding; of generating a second neural network that includes discrete edge weights from the neural network that includes precise edge weights by stochastic rounding; and of calculating an output by adding together the output of the first neural network and of the second neural network.

Abstract: A detective method, applied in a detective system comprising an activity-or-behavior model constructor, for activity-or-behavior model construction and automatic detection of activities of a subject system, comprises steps of using an unsupervised machine learning technique, a Natural Language Processing technique (NLP) and a supervised machine learning technique. As such, an activity-or-behavior model is built for predicting the future behaviors of the subject system and automatically detecting abnormal activities or behaviors of the subject system. The activity-or-behavior model is capable to handle multidimensional sensor data input from a plurality of sensor data streams and incorporate the sensor data values and a selected temporal information about at least one sensor data stream and between different sensor data streams.

Abstract: Disclosed embodiments provide a framework to assist customers in obtaining relevant responses from brands and other users to the intents communicated by these customers. In response to obtaining an intent, an intent messaging service identifies one or more users that can be provided with the intent to solicit responses to the intent. The one or more users are selected based on characteristics of the intent. The intent messaging service evaluates the responses to the intent from the one or more users to identify relevant responses that can be presented to the customer. The intent messaging service provides the relevant responses to the intent to the customer, which can determine which users to interact with to address the intent.

Abstract: Embodiments of the invention relate to a time-division multiplexed neurosynaptic module with implicit memory addressing for implementing a neural network. One embodiment comprises maintaining neuron attributes for multiple neurons and maintaining incoming firing events for different time steps. For each time step, incoming firing events for said time step are integrated in a time-division multiplexing manner. Incoming firing events are integrated based on the neuron attributes maintained. For each time step, the neuron attributes maintained are updated in parallel based on the integrated incoming firing events for said time step.

Abstract: In one implementation, a method of including a person in a CGR experience or excluding the person from the CGR experience is performed by a device including one or more processors, non-transitory memory, and a scene camera. The method includes, while presenting a CGR experience, capturing an image of scene; detecting, in the image of the scene, a person; and determining an identity of the person. The method includes determining, based on the identity of the person, whether to include the person in the CGR experience or exclude the person from the CGR experience. The method includes presenting the CGR experience based on the determination.

Abstract: A neural network training apparatus is described which has a network of worker nodes each having a memory storing a subgraph of a neural network to be trained. The apparatus has a control node connected to the network of worker nodes. The control node is configured to send training data instances into the network to trigger parallelized message passing operations which implement a training algorithm which trains the neural network. At least some of the message passing operations asynchronously update parameters of individual subgraphs of the neural network at the individual worker nodes.

Abstract: An intelligent monitoring device including a processor and an accelerometer and/or a device that includes at least one processor and at least one image sensor. The intelligent monitoring device is configured to observe at least one machine. The intelligent monitoring device is further configured to utilize its processor, or the processor of a coupled system, to recognize actions carried out on or by the at least one machine and infer the state of the machine. The intelligent monitoring device or the coupled system is further configured to provide alerts or help respond to queries about the status of the at least one machine. For example, the intelligent monitoring camera will infer the state of a washing machine based on its observations and provide an alert (optionally only to the user recognized to have loaded the washer) when the washing machine is ready to be unloaded.

Abstract: A system and method of animating an image of an object may include: receiving a first image, depicting a “puppet” object; sampling an input video, depicting a second, “driver” object, to obtain at least one second image; obtaining, by a first machine-learning (ML) model, a first identity-invariant feature of the puppet object, from the first image; obtaining at least one second identity-invariant feature of the driver object, from the respective at least one second image; calculating, by a second ML model, a transformation function, based on the first identity-invariant feature and the at least one second identity-invariant feature; applying the calculated transformation function on the first image, to produce one or more third images, depicting a target object, including at least one identity-invariant feature of the driver object; and appending the one or more third images to produce an output video depicting animation of the puppet object.

Abstract: Systems and methods that approximate and use branching operations on data encrypted by fully homomorphic encryption (FHE). The systems and methods may use polynomial approximation to convert “if” statements into “soft if” statements that may be applied to the FHE encrypted data in a manner that preserves the security of the systems and methods.

Abstract: Disclosed herein are techniques for obtain weights for neural network computations. In one embodiment, an integrated circuit may include memory configured to store a first weight and a second weight; a row of processing elements comprising a first processing element and a second processing element, the first processing element comprising a first weight register, the second processing element comprising a second weight register, both of the first weight register and the second weight register being controllable by a weight load signal; and a controller configured to: provide the first weight from the memory to the row of processing elements; set the weight load signal to enable the first weight to propagate through the row to reach the first processing element; and set the weight load signal to store the first weight at the first weight register and the flush value at the second weight register.

Abstract: A system for reconfiguring neural network architecture responsive to a system state is provided. A controller for modifying a neural network learning engine is configured to monitor a data stream having a data pattern by comparing the data pattern to a trained data pattern; identify a change in the data pattern of the data stream; determine a state of the neural network learning engine, the state defining one or more neural network parameters for monitoring the data stream with the neural network learning engine; and in response to identifying the change in the data pattern and determining the state, reconfigure an architectural configuration of the neural network learning engine by modifying the one or more neural network parameters.

Abstract: One embodiment relates to a neuromorphic network including electronic neurons and an interconnect circuit for interconnecting the neurons. The interconnect circuit includes synaptic devices for interconnecting the neurons via axon paths, dendrite paths and membrane paths. Each synaptic device includes a variable state resistor and a transistor device with a gate terminal, a source terminal and a drain terminal, wherein the drain terminal is connected in series with a first terminal of the variable state resistor. The source terminal of the transistor device is connected to an axon path, the gate terminal of the transistor device is connected to a membrane path and a second terminal of the variable state resistor is connected to a dendrite path, such that each synaptic device is coupled between a first axon path and a first dendrite path, and between a first membrane path and said first dendrite path.

Abstract: Disclosed herein are techniques for obtain weights for neural network computations. In one embodiment, an integrated circuit may include an arithmetic circuit configured to perform arithmetic operations for a neural network. The integrated circuit may also include a weight processing circuit configured to: acquire data from a memory device; receive configuration information indicating a size of each quantized weight of a set of quantized weights; extract the set of quantized weights from the data based on the size of the each weight indicated by the configuration information; perform de-quantization processing on the set of quantized weights to generate a set of de-quantized weights; and provide the set of de-quantized weights to the arithmetic circuit to enable the arithmetic circuit to perform the arithmetic operations. The memory device may be part of or external to the integrated circuit.

Abstract: In one aspect, this specification describes a recurrent neural network system implemented by one or more computers that is configured to process input sets to generate neural network outputs for each input set. The input set can be a collection of multiple inputs for which the recurrent neural network should generate the same neural network output regardless of the order in which the inputs are arranged in the collection. The recurrent neural network system can include a read neural network, a process neural network, and a write neural network. In another aspect, this specification describes a system implemented as computer programs on one or more computers in one or more locations that is configured to train a recurrent neural network that receives a neural network input and sequentially emits outputs to generate an output sequence for the neural network input.

Abstract: Provided are an application management method, a storage medium, and an electronic apparatus. The method includes: collecting a plurality of characteristic information of an application; learning the plurality of characteristic information to obtain a self-organizing neural network model of the application; calculating a first characteristic coefficient of the application; determining a second characteristic coefficient from the characteristic coefficient matrix according to the first characteristic coefficient; and judging whether the application can be cleaned up according to the second characteristic coefficient.

Abstract: A generative adversarial neural network (GAN) learns a particular task by being shown many examples. In one scenario, a GAN may be trained to generate new images including specific objects, such as human faces, bicycles, etc. Rather than training a complex GAN having a predetermined topology of features and interconnections between the features to learn the task, the topology of the GAN is modified as the GAN is trained for the task. The topology of the GAN may be simple in the beginning and become more complex as the GAN learns during the training, eventually evolving to match the predetermined topology of the complex GAN. In the beginning the GAN learns large-scale details for the task (bicycles have two wheels) and later, as the GAN becomes more complex, learns smaller details (the wheels have spokes).

Abstract: A discrete neural network is trained by training a neural network having an output layer so as to output discrete values. The output layer includes a plurality of nodes. Each node corresponding to one of a plurality of classes. The training includes activating the nodes by priority according to the corresponding class.

Abstract: An autonomous vehicle includes one or more sensors that measures the distance to and/or other properties of a target by illuminating the target with light. The sensor(s) provides sensor data indicative of one or more surface manifolds. Point data is generated with respect to the surface manifold(s). AHaH (Anti-Hebbian and Hebbian) based mechanism performs an AHaH-based feature extraction operation on the point data for compression and processing thereof.

Abstract: An electronic device is provided. The electronic device includes a memory storing recording data including a content of a conversation and at least one instruction, and a processor configured, by executing the at least one instruction, to input first data corresponding to a first voice in the content of the conversation into a first neural network model and acquire category information of the first data, and acquire category information of second data corresponding to a second voice in the content of the conversation. The processor is configured to, based on the category information of the first data and the category information of the second data being different, train the first neural network model based on the category information of the second data and the first data.

Abstract: Systems and methods of visual mining of user behavior patterns are disclosed. A plurality of clickstreams may be received. Each clickstream may represent a corresponding sequence of user actions. A visual representation of each clickstream may be caused to be displayed on a device. Each visual representation may comprise a distinct graphical element for each user action of the corresponding clickstream, and each visual representation may be configured to indicate a frequency level of the corresponding clickstream. Each distinct graphical element may comprise a geometric shape and a corresponding color that distinctly represents the corresponding user action. Graphical elements of each visual representation may comprise a size that is proportional to the frequency level of the corresponding clickstream. An indication of a selection of one of the visual representations may be received, and additional information about the corresponding clickstream may be caused to be displayed on the device.

Abstract: A neuromorphic device includes a pre-synaptic neuron, a synapse electrically coupled to the pre-synaptic neuron through a row line, and a post-synaptic neuron electrically coupled to the synapse through a column line. The post-synaptic neuron includes an integrator, a comparator, and an error corrector including an error detector and a correction signal generator. The comparator and the error corrector receive an output of the integrator.

Abstract: In one embodiment, a set of feature vectors can be derived from any biometric data, and then using a deep neural network (“DNN”) on those one-way homomorphic encryptions (i.e., each biometrics' feature vector) an authentication system can determine matches or execute searches on encrypted data. Each biometrics' feature vector can then be stored and/or used in conjunction with respective classifications, for use in subsequent comparisons without fear of compromising the original biometric data. In various embodiments, the original biometric data is discarded responsive to generating the encrypted values. In another embodiment, the homomorphic encryption enables computations and comparisons on cypher text without decryption of the encrypted feature vectors. Security of such privacy enable biometrics can be increased by implementing an assurance factor (e.g., liveness) to establish a submitted biometric has not been spoofed or faked.

Abstract: Methods, systems, and apparatus for selecting actions to be performed by an agent interacting with an environment. One system includes a high-level controller neural network, low-level controller network, and subsystem. The high-level controller neural network receives an input observation and processes the input observation to generate a high-level output defining a control signal for the low-level controller. The low-level controller neural network receives a designated component of an input observation and processes the designated component and an input control signal to generate a low-level output that defines an action to be performed by the agent in response to the input observation.

Abstract: Computer-implemented methods, systems, and apparatus, including computer-readable medium, for generating an action selection policy for causing an execution device to complete a task are described. Data representing a task that is divided into a sequence of subtasks are obtained. Data specifying a strategy neural network (SNN) for a subtask in the sequence of subtasks are obtained. The SNN receives inputs include a sequence of actions that reach an initial state of the subtask, and predicts an action selection policy of the execution device for the subtask. The SNN is trained based on a value neural network (VNN) for a next subtask that follows the subtask in the sequence of subtasks. An input to the SNN is determined. The input includes a sequence of actions that reach a subtask initial state of the subtask. An action selection policy for completing the subtask is determined based on an output of the SNN.

Abstract: A convolution neural network (CNN) model is trained and pruned at a pruning ratio. The model is then trained and pruned one or more times without constraining the model according to any previous pruning step. The pruning ratio may be increased at each iteration until a pruning target is reached. The model may then be trained again with pruned connections masked. The process of pruning, retraining, and adjusting the pruning ratio may also be repeated one or more times with a different pruning target.

Abstract: A method of updating a neural network may be provided. A method may include selecting a number of neurons for a layer for a neural network such that the number of neurons in the layer is less than at least one of a number of neurons in a first layer of the neural network and a number of neurons in a second, adjacent layer of the neural network. The method may further include and at least one of inserting the layer between the first layer and the second layer of the neural network and replacing one of the first layer and the second layer with the layer to reduce a number of connections in the neural network.

Abstract: Embodiments of the invention provide a neurosynaptic network circuit comprising multiple neurosynaptic devices including a plurality of neurosynaptic core circuits for processing one or more data packets. The neurosynaptic devices further include a routing system for routing the data packets between the core circuits. At least one of the neurosynaptic devices is faulty. The routing system is configured for selectively bypassing each faulty neurosynaptic device when processing and routing the data packets.

Abstract: A method for performing machine learning includes assigning processing jobs to a plurality of model learners, using a central parameter server. The processing jobs includes solving gradients based on a current set of parameters. As the results from the processing job are returned, the set of parameters is iterated. A degree of staleness of the solving of the second gradient is determined based on a difference between the set of parameters when the jobs are assigned and the set of parameters when the jobs are returned. The learning rates used to iterate the parameters based on the solved gradients are proportional to the determined degrees of staleness.

Abstract: We describe a system and a method that ascertains the strengths of links between pairs of biological sequence variants, by determining numerical link distances that measure the similarity of the molecular phenotypes of the variants. The link distances may be used to associate knowledge about labeled variants to other variants and to prioritize the other variants for subsequent analysis or interpretation. The molecular phenotypes are determined using a neural network, called a molecular phenotype neural network, and may include numerical or descriptive attributes, such as those describing protein-DNA interactions, protein-RNA interactions, protein-protein interactions, splicing patterns, polyadenylation patterns, and microRNA-RNA interactions. Linked genetic variants may be used to ascertain pathogenicity in genetic testing, to identify drug targets, to identify patients that respond similarly to a drug, to ascertain health risks, or to connect patients that have similar molecular phenotypes.

Abstract: An exemplary embodiment may describe a convolutional explainable neural network. A CNN-XNN may receive input, such as 2D or multi-dimensional data, a patient history, or any other relevant information. The input data is segmented into various objects and a knowledge encoding layer may identify and extract various features from the segmented objects. The features may be weighted. An output layer may provide predictions and explanations based on the previous layers. The explanation may be determined using a reverse indexing mechanism (Backmap). The explanation may be processed using a Kernel Labeler method that allows the labelling of the progressive refinement of patterns, symbols and concepts from any data format that allows a pattern recognition kernel to be defined allowing integration of neurosymbolic processing within CNN-XNNs. The optional addition of meta-data and causal logic allows for the integration of connectionist models with symbolic logic processing.

Abstract: A system for providing dynamic constitutional guidance. The system includes a label generator module configured to receive a periodic longevity factor, retrieve a user periodic longevity factor training set, and generate a naïve Bayes classification algorithm utilizing the user periodic longevity factor training set. The system includes a clustering module configured to receive a user adherence factor, retrieve a user adherence factor training set, and generate a k-means clustering algorithm using the user adherence factor training set. The system includes a processing module the processing module configured to retrieve a user ameliorative plan, evaluate a user ameliorative plan, generate an updated user ameliorative plan, and display the updated user ameliorative plan on a graphical user interface.

Abstract: Embodiments of the present invention describe a hierarchical cortical emulation using a scratchpad memory device and a storage class memory device. The scratchpad memory device is partitioned into a first subset of memory locations and a second subset of memory locations. A processor from a neural network device is assigned a first memory portion from the first subset, a second memory portion from the second subset, and a third memory portion from the storage class memory device. Further the neural network device and a memory controller perform a compute cycle for a hierarchical level k, 1?k?n, n being total number of levels. A compute cycle includes performing, by the processor, computations from the level k using neuron data stored in the first memory portion, and in parallel, copying by the memory controller, the neuron data for a hierarchical level k+1 from the third memory portion to the second memory portion.

Abstract: An information processing apparatus includes a control circuit configured to set or transmit a first learning parameter to a determining device that performs processing based on a learning parameter. The control circuit sets or transmits a second learning parameter instead of the first learning parameter to the determining device in a case where a result of a determination made by the determining device satisfies a predetermined condition. The first learning parameter is a learning parameter that is obtained by performing machine learning using a first learning data group. The second learning parameter is a learning parameter that is obtained by performing machine learning using a second learning data group. The first learning data group encompasses the second learning data group and includes learning data that is not included in the second learning data group.

Abstract: The present disclosure provides an approach for training a machine learning model by first training the model on a generic dataset and then iteratively training the model on “easy” domain specific training data before moving on to “difficult” domain specific training data. Inputs of a domain-specific dataset are run on the generically-trained model to determine which inputs generate an accuracy score above a threshold. The inputs with an accuracy score above a threshold are used to retrain the model, along with the corresponding outputs. The retraining continues until all domain specific dataset has been used to train the model, or until no remaining inputs of the domain specific dataset generate an accuracy score, when run on the model, that is above a threshold.

Abstract: A method of configuring a hardware implementation of a Convolutional Neural Network (CNN), the method comprising: determining, for each of a plurality of layers of the CNN, a first number format for representing weight values in the layer based upon a distribution of weight values for the layer, the first number format comprising a first integer of a first predetermined bit-length and a first exponent value that is fixed for the layer; determining, for each of a plurality of layers of the CNN, a second number format for representing data values in the layer based upon a distribution of expected data values for the layer, the second number format comprising a second integer of a second predetermined bit-length and a second exponent value that is fixed for the layer; and storing the determined number formats for use in configuring the hardware implementation of a CNN.

Abstract: System, methods, and other embodiments described herein relate to estimating ego-motion. In one embodiment, a method for estimating ego-motion based on a plurality of input images in a self-supervised system includes receiving a source image and a target image, determining a depth estimation Dt based on the target image, determining a depth estimation Ds based on a source image, and determining an ego-motion estimation in a form of a six degrees-of-freedom (6 DOF) transformation between the target image and the source image by inputting the depth estimations (Dt, Ds), the target image, and the source image into a two-stream network architecture trained to output the 6 DOF transformation based at least in part on the depth estimations (Dt, Ds), the target image, and the source image.

Abstract: Homeostasis-maintaining binary neural networks such as hierarchical temporal memories are provided. In various embodiments, a region of an artificial neural network is initialized. The region comprises a plurality of neurons and has a permanence value associated with each potential synaptic connection between neurons. The initialization comprises connecting a subset of the potential synaptic connections by synapses. A plurality of time-ordered inputs to the region is received. Some of the plurality of neurons are thereby caused to fire upon receipt of each time-ordered input. Upon receipt of each time-ordered input, for each potential synaptic connection between neurons, the permanence value is adjusted according to a firing sequence of the plurality of neurons. Those potential synaptic connections having a permanence value above a predetermined permanence threshold are connected when the total number of connected synapses in the region does not exceed a predetermined connectivity threshold.

Abstract: Embodiments are directed to a method for accelerating machine learning using a plurality of graphics processing units (GPUs), involving receiving data for a graph to generate a plurality of random samples, and distributing the random samples across a plurality of GPUs. The method may comprise determining a plurality of communities from the random samples using unsupervised learning performed by each GPU. A plurality of sample groups may be generated from the communities and may be distributed across the GPUs, wherein each GPU merges communities in each sample group by converging to an optimal degree of similarity. In addition, the method may also comprise generating from the merged communities a plurality of subgraphs, dividing each sub-graph into a plurality of overlapping clusters, distributing the plurality of overlapping clusters across the plurality of GPUs, and scoring each cluster in the plurality of overlapping clusters to train an AI model.

Abstract: A method for hardware-implemented training of a feedforward artificial neural network is provided. The method comprises: generating a first output signal by processing an input signal with the network, wherein a cost quantity assumes a first cost value; measuring the first cost value; defining a group of at least one synaptic weight of the network for variation; varying each weight of the group by a predefined weight difference; after the variation, generating a second output signal from the input signal to measure a second cost value; comparing the first and second cost values; and determining, based on the comparison, a desired weight change for each weight of the group such that the cost function does not increase if the respective desired weight changes are added to the weights of the group. The desired weight change is based on the weight difference times ?1, 0, or +1.

Abstract: A method for implementing spiking neural network computations, the method including defining a dynamic node response function that exhibits spikes, where spikes are temporal nonlinearities for representing state over time; defining a static representation of said node response function; and using the static representation of the node response function to train a neural network. A system for implementing the method is also disclosed.

Abstract: A neuromorphic device may include a pre-synaptic neuron, a row line extending from the pre-synaptic neuron in a row direction, a post-synaptic neuron, a column line extending from the post-synaptic neuron in a column direction, and a synapse coupled between the row line and the column line. The synapse may be disposed in an intersection region between the row line and the column line. The post-synaptic neuron may include a subtracting circuit.

Abstract: A computer-implemented method includes receiving, by a computing device, input activations and determining, by a controller of the computing device, whether each of the input activations has either a zero value or a non-zero value. The method further includes storing, in a memory bank of the computing device, at least one of the input activations. Storing the at least one input activation includes generating an index comprising one or more memory address locations that have input activation values that are non-zero values. The method still further includes providing, by the controller and from the memory bank, at least one input activation onto a data bus that is accessible by one or more units of a computational array. The activations are provided, at least in part, from a memory address location associated with the index.

Abstract: Methods and systems are disclosed using improved Convolutional Neural Networks (CNN) for image processing. In one example, an input image is down-sampled into smaller images with a smaller resolution than the input image. The down-sampled smaller images are processed by a CNN having a last layer with a reduced number of nodes than a last layer of a full CNN used to process the input image at a full resolution. A result is outputted based on the processed down-sampled smaller images by the CNN having a last layer with a reduced number of nodes. In another example, shallow CNN networks are built randomly. The randomly built shallow CNN networks are combined to imitate a trained deep neural network (DNN).