APPARATUS, METHOD, DEVICE AND MEDIUM FOR LOSS BALANCING IN MULTI-TASK LEARNING

    公开(公告)号:US20240303485A1

    公开(公告)日:2024-09-12

    申请号:US18571616

    申请日:2021-12-02

    CPC classification number: G06N3/08

    Abstract: The disclosure provides an apparatus, method, device, and medium for loss balancing in MTL. The apparatus includes interface circuitry and processor circuitry. The processor circuitry is configured to initialize parameters of shared layers of a deep neural network for MTL using a pre-trained neural network; determine a custom interval consisting of a designated number of mini-batch training steps and a designated window of N custom intervals (N>2); for each task, calculate a loss change rate between each pair of N−1 pairs of neighboring custom intervals within a designated window prior to a present custom interval and a gradient magnitude with respect to selected shared weights within the designated window prior to the present custom interval, and adjust, a weight of the task, based on the calculated loss change rate and gradient magnitude with respect to selected shared weights.

    TECHNIQUES FOR DETERMINING A CURRENT LOCATION OF A MOBILE DEVICE

    公开(公告)号:US20190094027A1

    公开(公告)日:2019-03-28

    申请号:US16081129

    申请日:2016-03-30

    Abstract: Various embodiments are directed to techniques for determining a current location of a mobile device. An apparatus includes a SLAM candidate component to identify a first candidate key frame matching a current captured frame by a first degree from an interval-based key frame set with key frames selected on a recurring interval from multiple earlier captured frames captured by mobile device camera of surroundings within a defined area, a CNN candidate component to identify a second candidate key frame matching the current captured frame by a second degree from a difference-based key frame set with key frames selected from the multiple earlier captured frames based on a degree of difference from all key frames already therein, and a position estimation component to determine a current location of the mobile device from estimates of differences between the current location and locations of the first and second candidate key frames.

Patent Agency Ranking