Significant progress has been made with deep neural networks recently. Sharing trained models of deep neural networks has been a very important in the rapid progress of research and development of these systems. At the same time, it is necessary to protect the rights to shared trained models. To this end, we propose to use digital watermarking technology to protect intellectual property and detect intellectual property infringement in the use of trained models. First, we formulate a new problem: embedding watermarks into deep neural networks. We also define requirements, embedding situations, and attack types on watermarking in deep neural networks. Second, we propose a general framework for embedding a watermark in model parameters, using a parameter regularizer. Our approach does not impair the performance of networks into which a watermark is placed because the watermark is embedded while training the host network. Finally, we perform comprehensive experiments to reveal the potential of watermarking deep neural networks as the basis of this new research effort. We show that our framework can embed a watermark during the training of a deep neural network from scratch, and during fine-tuning and distilling, without impairing its performance. The embedded watermark does not disappear even after fine-tuning or parameter pruning; the watermark remains complete even after 65% of parameters are pruned.

Although deep neural networks have made tremendous progress in the area of multimedia representation, training neural models requires a large amount of data and time. It is well-known that utilizing trained models as initial weights often achieves lower training error than neural networks that are not pre-trained. A fine-tuning step helps to reduce both the computational cost and improve performance. Therefore, sharing trained models has been very important for the rapid progress of research and development. In addition, trained models could be important assets for the owner(s) who trained them, hence we regard trained models as intellectual property. In this paper, we propose a digital watermarking technology for ownership authorization of deep neural networks. First, we formulate a new problem: embedding watermarks into deep neural networks. We also define requirements, embedding situations, and attack types on watermarking in deep neural networks. Second, we propose a general framework for embedding a watermark in model parameters, using a parameter regularizer. Our ap- S. Satoh National Institute of Informatics 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo, 101-8430, Japan E-mail: satoh@nii.ac.jp proach does not impair the performance of networks into which a watermark is placed because the watermark is embedded while training the host network. Finally, we perform comprehensive experiments to reveal the potential of watermarking deep neural networks as the basis of this new research effort. We show that our framework can embed a watermark during the training of a deep neural network from scratch, and during fine-tuning and distilling, without impairing its performance. The embedded watermark does not disappear even after fine-tuning or parameter pruning; the watermark remains complete even after 65% of parameters are pruned.

In the present review, how to measure motor imagery ability, brain activity during motor imagery, the benefits of motor imagery practice, and the influence of sensory inputs on motor imagery, are summarized. First, the classification of motor imagery is explained. Many methods have been utilized to evaluate motor imagery ability. For example, questionnaires, mental chronometry, and mental rotation tasks have been used in the psychological approach. Brain activity has been measured utilizing transcranial magnetic stimulation (TMS), functional magnetic resonance imaging (fMRI), positron emission tomography (PET), and electroencephalography (EEG). Some brain regions are activated motor execution in both and motor imagery, including the supplementary motor area (SMA), the premotor cortex (PM) and the parietal cortex. Although motor imagery is done without movement or muscle contraction, sensory input from the periphery interacts with motor imagery. Brain activation during imagery of an action, as assessed by TMS, is stronger when sensory inputs resemble those present during the actual execution of the action. Many studies have provided evidence of the effects of motor imagery practice on basic motor skills and sport performance. Most elite athletes (70-90%) report that they use motor imagery to improve performance, and professional players, as compared to amateurs, utilize imagery practice more often. Many studies have confirmed that motor imagery practice can also be useful not only in sports, but also for improving performance in patient rehabilitation programs.

Dynamic visual acuity (DVA) is defined as the ability to discriminate the fine parts of a moving object. DVA is generally better in athletes than in non-athletes, and the better DVA of athletes has been attributed to a better ability to track moving objects. In the present study, we hypothesized that the better DVA of athletes is partly derived from better perception of moving images on the retina through some kind of perceptual learning. To test this hypothesis, we quantitatively measured DVA in baseball players and non-athletes using moving Landolt rings in two conditions. In the first experiment, the participants were allowed to move their eyes (free-eye-movement conditions), whereas in the second they were required to fixate on a fixation target (fixation conditions). The athletes displayed significantly better DVA than the non-athletes in the free-eye-movement conditions. However, there was no significant difference between the groups in the fixation conditions. These results suggest that the better DVA of athletes is primarily due to an improved ability to track moving targets with their eyes, rather than to improved perception of moving images on the retina.

These results suggest that the better DVA of baseball players is primarily due to a better ability to track moving objects with their eyes rather than to improved perception of moving images on the retina. This skill is probably obtained through baseball training.