首页> 外国专利> Deep learning based techniques for pre training deep convolution neural networks

Deep learning based techniques for pre training deep convolution neural networks

机译:基于深入训练深度卷积神经网络的基于深度学习技术

摘要

The disclosed techniques include systems and methods for reducing the overfitting of neural network implementation models to handle an array of amino acids and associated positional specific frequency matrices.The system generates a benign and labelled supplemental training example sequence pair comprising an arrangement that moves the target amino acid position from the starting position to the end position.The complementary sequence pairs complement the virulence or benign missense training example sequence pairs.This has the same amino acid in the amino acid reference and alternative sequences.The system comprises logic for inputting the same supplemental training position specific frequency matrix (PFM) as a benign or pathogenic missense PFM at a matching start and end position along with each complementary array pair.The system comprises logic for weakening the effects of training PFM training during training the neural network implementation model by including supplementary training example PFM in training example data.
机译:所公开的技术包括用于减少神经网络实现模型的过度接收以处理氨基酸阵列和相关位置特定频率矩阵的系统和方法。系统产生良性和标记的补充训练示例序列对,该序列对包括移动目标氨基的布置从起始位置到最终位置的酸位置。互补序列对补充了毒力或良性畸形训练示例序列对。该系统在氨基酸参考和替代序列中具有相同的氨基酸。该系统包括用于输入相同补充的逻辑训练位置特定频率矩阵(PFM)作为匹配开始和结束位置的良性或致病畸形PFM以及每个互补阵列对。该系统包括通过包括在培训神经网络实现模型期间培训培训PFM训练训练的效果的逻辑补充火车训练示例数据中的示例PFM。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号