A three-stage learning algorithm for deep multilayer perceptron (DMLP) with effective weight initialisation based on sparse auto-encoder is proposed in this paper, which aims to overcome difficulties in training deep neural networks with limited training data in high-dimensional feature space. At the first stage, unsupervised learning is adopted using sparse auto-encoder to obtain the initial weights of the feature extraction layers of the DMLP. At the second stage, error back-propagation is used to train the DMLP by fixing the weights obtained at the first stage for its feature extraction layers. At the third stage, all the weights of the DMLP obtained at the second stage are refined by error back-propagation. Network structures and values of learning parameters are determined through cross-validation, and test datasets unseen in the cross-validation are used to evaluate the performance of the DMLP trained using the three-stage learning algorithm. Experimental results show that the proposed method is effective in combating overfitting in training deep neural networks.
Some maps of the chaotic firefly algorithm were selected to select variables for data on blood diseases and blood vessels obtained from Nasiriyah General Hospital where the data were tested and tracking the distribution of Gamma and it was concluded that a Chebyshevmap method is more efficient than a Sinusoidal map method through mean square error criterion.
A multidimensional systolic arrays realization of LMS algorithm by a method of mapping regular algorithm onto processor array, are designed. They are based on appropriately selected 1-D systolic array filter that depends on the inner product sum systolic implementation. Various arrays may be derived that exhibit a regular arrangement of the cells (processors) and local interconnection pattern, which are important for VLSI implementation. It reduces latency time and increases the throughput rate in comparison to classical 1-D systolic arrays. The 3-D multilayered array consists of 2-D layers, which are connected with each other only by edges. Such arrays for LMS-based adaptive (FIR) filter may be opposed the fundamental requirements of fa
... Show MoreHuman posture estimation is a crucial topic in the computer vision field and has become a hotspot for research in many human behaviors related work. Human pose estimation can be understood as the human key point recognition and connection problem. The paper presents an optimized symmetric spatial transformation network designed to connect with single-person pose estimation network to propose high-quality human target frames from inaccurate human bounding boxes, and introduces parametric pose non-maximal suppression to eliminate redundant pose estimation, and applies an elimination rule to eliminate similar pose to obtain unique human pose estimation results. The exploratory outcomes demonstrate the way that the proposed technique can pre
... Show More