Language processing reveals conceptual combination in human infants

Across three eye-tracking experiments, we taught 12-month-olds (N = 60) two novel quantity labels denoting sets of one and two (e.g., “mize” for 1; “padu” for 2). We then showed that they could not only generalize these labels to sets of previously unseen objects, but also combine them with familiar category labels acquired prior to the lab visit (e.g., “ball”, “duck”). Their eye movements revealed adult-like compositional procedures that go beyond serial processing of constituent meanings. These findings indicate that certain combinatorial processes involved in extracting complex linguistic meaning are already available by the end of the first year of life and are ready to support language comprehension.

T-MIS: Transparency Adaptation in Medical Image Segmentation

We often locate ourselves in a trade-off situation between what is predicted and understanding why the predictive modeling made such a prediction. This high-risk medical segmentation task is no different where we try to interpret how well has the model learned from the image features irrespective of its accuracy. We propose image-specific fine-tuning to make a deep learning model adaptive to specific medical imaging tasks. Experimental results reveal that: a) proposed model is more robust to segment previously unseen objects (negative test dataset) than state-of-the-art CNNs; b) image-specific fine-tuning with the proposed heuristics significantly enhances segmentation accuracy; and c) our model leads to accurate results with fewer user interactions and less user time than conventional interactive segmentation methods. The model successfully classified ’no polyp’ or ’no instruments’ in the image irrespective of the absence of negative data in training samples from Kvasir-seg and Kvasir-Instrument datasets.

RF-ray

This paper presents a non-invasive design, namely RF-ray, to recognize the shape and material of an object simultaneously. RF-ray puts the object approximate to an RFID tag array, and explores the propagation effect as well as coupling effect between RFIDs and the object for sensing. In contrast to prior proposals, RF-ray is capable to recognize unseen objects, including unseen shape-material pairs and unseen materials within a certain container. To make it real, RF-ray introduces a sensing capability enhancement module and leverages a two-branch neural network for shape profiling and material identification respectively. Furthermore, we incorporate a Zero-Shot Learning based embedding module that incorporates the well-learned linguistic features to generalize RF-ray to recognize unseen materials. We build a prototype of RF-ray using commodity RFID devices. Comprehensive real-world experiments demonstrate our system can achieve high object recognition performance.

Unsupervised Novelty Characterization in Physical Environments Using Qualitative Spatial Relations

Detecting, characterizing and adapting to novelty, whether in the form of previously unseen objects or phenomena, or unexpected changes in the behavior of known elements, is essential for Artificial Intelligence agents to operate reliably in unconstrained real-world environments. We propose an automatic, unsupervised approach to novelty characterization for dynamic domains, based on describing the behaviors and interactions of objects in terms of their possible actions. To abstract from the variety of realizations of an action that can occur in physical domains, we model states in terms of qualitative spatial relations (QSRs) between their entities. By first learning a model of actions in the non-novel environment from the state transitions observed as the agent interacts with the world, we can detect novelty by the persistent deviations from this model that it causes, and characterize the novelty by new or modified actions. We also present a new method of learning action models from observation, based on conceptual similarity and hierarchical clustering.

Deep Reinforcement Learning Enhanced Convolutional Neural Networks for Robotic Grasping

Robotic grasping is an important task for various industrial applications. However, combining detecting and grasping to perform a dynamic and efficient object moving is still a challenge for robotic grasping. Meanwhile, it is time consuming for robotic algorithm training and testing in realistic. Here we present a framework for dynamic robotic grasping based on deep Q-network (DQN) in a virtual grasping space. The proposed dynamic robotic grasping framework mainly consists of the DQN, the convolutional neural network (CNN), and the virtual model of robotic grasping. After observing the result generated by applying the generative grasping convolutional neural network (GG-CNN), a robotic manipulation conducts actions according to Q-network. Different actions generate different rewards, which are implemented to update the neural network through loss function. The goal of this method is to find a reasonable strategy to optimize the total reward and finally accomplish a dynamic grasping process. In the test of virtual space, we achieve an 85.5% grasp success rate on a set of previously unseen objects, which demonstrates the accuracy of DQN enhanced GG-CNN model. The experimental results show that the DQN can efficiently enhance the GG-CNN by considering the grasping procedure (i.e. the grasping time and the gripper’s posture), which makes the grasping procedure stable and increases the success rate of robotic grasping.

Partial caging: a clearance-based definition, datasets, and deep learning

Caging grasps limit the mobility of an object to a bounded component of configuration space. We introduce a notion of partial cage quality based on maximal clearance of an escaping path. As computing this is a computationally demanding task even in a two-dimensional scenario, we propose a deep learning approach. We design two convolutional neural networks and construct a pipeline for real-time planar partial cage quality estimation directly from 2D images of object models and planar caging tools. One neural network, CageMaskNN, is used to identify caging tool locations that can support partial cages, while a second network that we call CageClearanceNN is trained to predict the quality of those configurations. A partial caging dataset of 3811 images of objects and more than 19 million caging tool configurations is used to train and evaluate these networks on previously unseen objects and caging tool configurations. Experiments show that evaluation of a given configuration on a GeForce GTX 1080 GPU takes less than 6 ms. Furthermore, an additional dataset focused on grasp-relevant configurations is curated and consists of 772 objects with 3.7 million configurations. We also use this dataset for 2D Cage acquisition on novel objects. We study how network performance depends on the datasets, as well as how to efficiently deal with unevenly distributed training data. In further analysis, we show that the evaluation pipeline can approximately identify connected regions of successful caging tool placements and we evaluate the continuity of the cage quality score evaluation along caging tool trajectories. Influence of disturbances is investigated and quantitative results are provided.

Informative neural representations of unseen objects during higher-order processing in human brains and deep artificial networks

Despite advances in the neuroscience of visual consciousness over the last decades, we still lack a framework for understanding the scope of unconscious processing and how it relates to conscious experience. Previous research observed brain signatures of unconscious contents in visual cortex, but these have not been identified in a reliable manner, with low trial numbers and signal detection theoretic constraints not allowing to decisively discard conscious perception. Critically, the extent to which unconscious content is represented in high-level processing stages along the ventral visual stream and linked prefrontal areas remains unknown. Using a within-subject, high-precision, highly-sampled fMRI approach, we show that unconscious contents, even those associated with null sensitivity, can be reliably decoded from multivoxel patterns that are highly distributed along the ventral visual pathway and also involving prefrontal substrates. Notably, the neural representation in these areas generalised across conscious and unconscious visual processing states, placing constraints on prior findings that fronto-parietal substrates support the representation of conscious contents and suggesting revisions to models of consciousness such as the neuronal global workspace. We then provide a computational model simulation of visual information processing/representation in the absence of perceptual sensitivity by using feedforward convolutional neural networks trained to perform a similar visual task to the human observers. The work provides a novel framework for pinpointing the neural representation of unconscious knowledge across different task domains.

Semantic priming supports infants’ ability to learn names of unseen objects

Human language permits us to call to mind objects, events, and ideas that we cannot witness directly. This capacity rests upon abstract verbal reference: the appreciation that words are linked to mental representations that can be established, retrieved and modified, even when the entities to which a word refers is perceptually unavailable. Although establishing verbal reference is a pivotal achievement, questions concerning its developmental origins remain. To address this gap, we investigate infants’ ability to establish a representation of an object, hidden from view, from language input alone. In two experiments, 15-month-olds (N = 72) and 12-month-olds (N = 72) watch as an actor names three familiar, visible objects; she then provides a novel name for a fourth, hidden fully from infants’ view. In the Semantic Priming condition, the visible familiar objects all belong to the same semantic neighborhood (e.g., apple, banana, orange). In the No Priming condition, the objects are drawn from different semantic neighborhoods (e.g., apple, shoe, car). At test infants view two objects. If infants can use the naming information alone to identify the likely referent, then infants in the Semantic Priming, but not in the No Priming condition, will successfully infer the referent of the fourth (hidden) object. Brief summary of results here. Implications for the development of abstract verbal reference will be discussed.