Abstract: A method for using a multi-hop reasoning framework to perform multi-step compositional long-term reasoning is presented. The method includes extracting feature maps and frame-level representations from a video stream by using a convolutional neural network (CNN), performing object representation learning and detection, linking objects through time via tracking to generate object tracks and image feature tracks, feeding the object tracks and the image feature tracks to a multi-hop transformer that hops over frames in the video stream while concurrently attending to one or more of the objects in the video stream until the multi-hop transformer arrives at a correct answer, and employing video representation learning and recognition from the objects and image context to locate a target object within the video stream.

Abstract: A computer-implemented method is provided for compositional reasoning. The method includes producing a set of primitive predictions from an input sequence. Each of the primitive predictions is of a single action of a tracked subject to be composed in a complex action comprising multiple single actions. The method further includes performing contextual rule filtering of the primitive predictions to pass through filtered primitive predictions that interact with one or more entities of interest in the input sequence with respect to predefined contextual interaction criteria. The method includes performing, by a processor device, temporal rule matching by matching the filtered primitive predictions according to pre-defined temporal rules to identify complex event patterns in the sequence of primitive predictions.

Abstract: A method is provided for determining entailment between an input premise and an input hypothesis of different modalities. The method includes extracting features from the input hypothesis and an entirety of and regions of interest in the input premise. The method further includes deriving intra-modal relevant information while suppressing intra-modal irrelevant information, based on intra-modal interactions between elementary ones of the features of the input hypothesis and between elementary ones of the features of the input premise. The method also includes attaching cross-modal relevant information to the features from the input premise to the features from the input hypothesis to form a cross-modal representation, based on cross-modal interactions between pairs of different elementary features from different modalities.

Abstract: A computer-implemented method and system are provided for teaching syntax for training a neural network based natural language inference model. The method includes selectively performing, by the hardware processor, person reversal on a set of hypothesis sentences, based on person reversal prevention criteria, to obtain a first training data set. The method further includes enhancing, by the hardware processor, a robustness of the neural network based natural language inference model to syntax changes by training the neural network based natural language inference model on original training data combined with the first data set.

Abstract: Systems and methods for building a distributed learning framework, including generating a sparse communication network graph with a high overall spectral gap. The generating includes computing model parameters in distributed shared memory of a cluster of a plurality of worker nodes; determining a spectral gap of an adjacency matrix for the cluster using a stochastic reduce convergence analysis, wherein a spectral reduce is performed using a sparse reduce graph with a highest possible spectral gap value for a given network bandwidth capability; and optimizing the communication graph by iteratively performing the computing and determining until a threshold condition is reached. Each of the plurality of worker nodes is controlled using tunable approximation based on available bandwidth in a network in accordance with the generated sparse communication network graph.

Abstract: A context-aware attention-based neural network is provided for answering an input question given a set of purportedly supporting statements for the input question. The neural network includes a processing element. The processing element is configured to calculate a question representation for the input question, based on word annotations and word-level attentions calculated for the input question. The processing element is further configured to calculate a sentence representation for each of the purportedly supporting statements, based on word annotations and word-level attentions calculated for each of the purportedly supporting statements. The processing element is also configured to calculate a context representation for the set of purportedly supporting statements with respect to the sentence representation for each of the purportedly supporting statements.

Abstract: A computer-implemented method is provided for disentangled data generation. The method includes accessing, by a variational autoencoder, a plurality of supervision signals. The method further includes accessing, by the variational autoencoder, a plurality of auxiliary tasks that utilize the supervision signals as reward signals to learn a disentangled representation. The method also includes training the variational autoencoder to disentangle a sequential data input into a time-invariant factor and a time-varying factor using a self-supervised training approach which is based on outputs of the auxiliary tasks obtained by using the supervision signals to accomplish the plurality of auxiliary tasks.

Abstract: Aspects of the present disclosure describe systems, methods and structures including a network that recognizes action(s) from learned relationship(s) between various objects in video(s). Interaction(s) of objects over space and time is learned from a series of frames of the video. Object-like representations are learned directly from various 2D CNN layers by capturing the 2D CNN channels, resizing them to an appropriate dimension and then providing them to a transformer network that learns higher-order relationship(s) between them. To effectively learn object-like representations, we 1) combine channels from a first and last convolutional layer in the 2D CNN, and 2) optionally cluster the channel (feature map) representations so that channels representing the same object type are grouped together.

Abstract: Aspects of the present disclosure describe systems, methods and structures providing contextual grounding—a higher-order interaction technique to capture corresponding context between text entities and visual objects.

Abstract: Aspects of the present disclosure describe systems, methods, and structures that provide action recognition with high-order interaction with spatio-temporal object tracking. Image and object features are organized into into tracks, which advantageously facilitates many possible learnable embeddings and intra/inter-track interaction(s). Operationally, our systems, method, and structures according to the present disclosure employ an efficient high-order interaction model to learn embeddings and intra/inter object track interaction across the space and time for AR. Each frame is detected by an object detector to locate visual objects. Those objects are linked through time to form object tracks. The object tracks are then organized and combined with the embeddings as the input to our model. The model is trained to generate representative embeddings and discriminative video features through high-order interaction which is formulated as an efficient matrix operation without iterative processing delay.

Abstract: Aspects of the present disclosure describe systems, methods and structures for an efficient multi-person posetracking method that advantageously achieves state-of-the-art performance on PoseTrack datasets by only using keypoint information in a tracking step without optical flow or convolution routines. As a consequence, our method has fewer parameters and FLOPs and achieves faster FPS. Our method benefits from our parameter-free tracking method that outperforms commonly used bounding box propagation in top-down methods. Finally, we disclose tokenization and embedding multi-person pose keypoint information in the transformer architecture that can be re-used for other pose tasks such as pose-based action recognition.

Abstract: Security systems and methods for detecting intrusion events include one or more sensors configured to monitor an environment. A pruned convolutional neural network (CNN) is configured process information from the one or more sensors to classify events in the monitored environment. CNN filters having the smallest summed weights have been pruned from the pruned CNN. An alert module is configured to detect an intrusion event in the monitored environment based on event classifications. A control module is configured to perform a security action based on the detection of an intrusion event.

Abstract: A method for performing question answer (QA) tasks that includes entering an input into an encoder portion of an adaptive memory network, wherein the encoder portion parses the input into entities of text for arrangement of memory banks. A bank controller of the adaptive memory network organizes the entities into progressively weighted banks within the arrangement of memory banks. The arrangement of memory banks may be arranged to have an initial memory bank having lowest relevance for lowest relevance entities being closest to the encoder, and a final memory bank having a highest relevance for highest relevance entities being closes to a decoder. The method may continue with inferring an answer for the question answer (QA) task with the decoder analyzing only the highest relevance entities in the final memory bank.

Abstract: Methods and systems for pruning a convolutional neural network (CNN) include calculating a sum of weights for each filter in a layer of the CNN. The filters in the layer are sorted by respective sums of weights. A set of m filters with the smallest sums of weights is filtered to decrease a computational cost of operating the CNN. The pruned CNN is retrained to repair accuracy loss that results from pruning the filters.

Abstract: Systems and methods for predicting changes to an environment, including a plurality of remote sensors, each remote sensor being configured to capture images of an environment. A processing device is included on each remote sensor, the processing device configured to recognize and predict a change to the environment using a pruned convolutional neural network (CNN) stored on the processing device, the pruned CNN being trained to recognize features in the environment by training a CNN with a dataset and removing filters from layers of the CNN that are below a significance threshold for image recognition to produce the pruned CNN. A transmitter is configured to transmit the recognized and predicted change to a notification device such that an operator is alerted to the change.

Abstract: A computer-implemented method executed by at least one processor for performing mini-batching in deep learning by improving cache utilization is presented. The method includes temporally localizing a candidate clip in a video stream based on a natural language query, encoding a state, via a state processing module, into a joint visual and linguistic representation, feeding the joint visual and linguistic representation into a policy learning module, wherein the policy learning module employs a deep learning network to selectively extract features for select frames for video-text analysis and includes a fully connected linear layer and a long short-term memory (LSTM), outputting a value function from the LSTM, generating an action policy based on the encoded state, wherein the action policy is a probabilistic distribution over a plurality of possible actions given the encoded state, and rewarding policy actions that return clips matching the natural language query.

Abstract: Systems and methods for surveillance are described, including an image capture device configured to mounted to an autonomous vehicle, the image capture device including an image sensor. A storage device is included in communication with the processing system, the storage device including a pruned convolutional neural network (CNN) being trained to recognize obstacles in a road according to images captured by the image sensor by training a CNN with a dataset and removing filters from layers of the CNN that are below a significance threshold for image recognition to produce the pruned CNN. A processing device is configured to recognize the obstacles by analyzing the images captured by the image sensor with the pruned CNN and to predict movement of the obstacles such that the autonomous vehicle automatically and proactively avoids the obstacle according to the recognized obstacle and predicted movement.

Abstract: A machine learning method includes installing a plurality of model replicas for training on a plurality of computer learning nodes; receiving training data at a each model replica and updating parameters for the model replica after trailing; sending the parameters to other model replicas with a communication batch size; evaluating received parameters from other model replicas; and dynamically adjusting the communication batch size to balance computation and communication overhead and ensuring convergence even with a mismatch in processing abilities on different computer learning nodes.

Abstract: A method is provided for determining entailment between an input premise and an input hypothesis of different modalities. The method includes extracting features from the input hypothesis and an entirety of and regions of interest in the input premise. The method further includes deriving intra-modal relevant information while suppressing intra-modal irrelevant information, based on intra-modal interactions between elementary ones of the features of the input hypothesis and between elementary ones of the features of the input premise. The method also includes attaching cross-modal relevant information to the features from the input premise to the features from the input hypothesis to form a cross-modal representation, based on cross-modal interactions between pairs of different elementary features from different modalities.

Abstract: A computer-implemented method and system are provided for teaching syntax for training a neural network based natural language inference model. The method includes selectively performing, by the hardware processor, person reversal on a set of hypothesis sentences, based on person reversal prevention criteria, to obtain a first training data set. The method further includes enhancing, by the hardware processor, a robustness of the neural network based natural language inference model to syntax changes by training the neural network based natural language inference model on original training data combined with the first data set.