Tag:dimension

  • FTRL-Proximal

    Time:2020-5-25

    FTRL-ProximalFull nameFollowthe-Regularized-Leader Proximal, proposed by GoogleOnline learningAlgorithm, in dealing with regular terms with non smooth (for example, $l_ 1 $norm) is excellent in convex optimization. The traditional algorithm based on batch can not effectively deal with large-scale data and online data. However, many Internet applications, such as advertising, data comes one by one. For each […]

  • NGX of current limiting in access layer_ http_ limit_ req_ module

    Time:2020-5-21

    [please indicate the source of reprint]: https://segmentfault.com/a/1190000022599606 ngx_ http_ limit_ req_ Module module is a request flow limiting module based on the leaky bucket algorithm provided by nginx, which is used to limit the flow of requests corresponding to the specified key, such as limiting the request rate according to the IP dimension. ngx_ http_ […]

  • Tensorflow realizes simple CNN

    Time:2020-5-3

    The code of implementing a simple CNN using tensroflow is as follows for reference only #!/usr/bin/env python # -*- coding: utf-8 -*- import numpy as np import tensorflow as tf #download mnist datasets #55000 * 28 * 28 55000image from tensorflow.examples.tutorials.mnist import input_data mnist=input_data.read_data_sets(‘mnist_data’,one_hot=True) #one_hot is encoding format #The first dimension of none means sensor […]

  • The usage of function shape in numpy

    Time:2020-4-27

    Shape function is a function in numpy.core.fromnumeraric. Its function is to read the length of matrix. For example, shape [0] is to read the length of the first dimension of matrix. Its input parameters can make an integer represent a dimension or a matrix. So you may not understand. Let’s use various examples to illustrate […]

  • On batch reading of dataset image and detailed operation of dimension in tensorflow

    Time:2020-4-24

    Three dimensional reading pictures (W, h, c): import tensorflow as tf import glob import os def _parse_function(filename): # print(filename) image_string = tf.read_file(filename) image_decoded = tf.image.decode_image(image_string) # (375, 500, 3) image_resized = tf.image.resize_image_with_crop_or_pad(image_decoded, 200, 200) return image_resized with tf.Session() as sess: print( sess.run( img ).shape ) Read pictures of batch pictures (B, W, h, c): import […]

  • “Performance tuning” combo that programmers must practice [1]

    Time:2020-4-23

    In the article “performance tuning XYZ that programmers must master”, veteran brother explains why programmers need to cross the performance tuning level when they are developing towards architects based on their personal experience. This is an opportunity for us to establish a process, structured and systematic thinking. In addition, veteran brother also introduced the idea […]

  • Dropout in tensorflow

    Time:2020-4-22

    Hinton in the paper《Improving neural networks by preventing co-adaptation of feature detectors》It is proposed inDropout。DropoutIt is used to prevent over fitting of neural network. Dropout can be implemented in tensorflow in the following 3 ways. tf.nn.dropout def dropout(x, keep_prob, noise_shape=None, seed=None, name=None): Among them,xIs a float type sensor,keep_probIt is a floating-point scalar with a range […]

  • “Performance tuning” combo that programmers must practice [3]

    Time:2020-4-21

    Index of articles before performance tuning series: Performance tuning that programmers must master:Veteran brother explains why programmers need to cross the performance optimization level when they are developing towards architects, and introduces the idea of optimizing performance from X, y and Z dimensions. Optimize system performance from the X dimension:Veteran brother shared the idea of […]

  • Summary of spark OLAP higher order analysis functions

    Time:2020-4-20

    We are often confused that SQL can’t be written in data mining and report analysis scenarios, or because SQL is too long to be readable. Today, I summarized some high-order functions in spark SQL, which will help your business and improve your work efficiency 100 times GROUPING__ID,CUBE,ROLLUP It can quickly realize multi-dimensional free combination analysis […]

  • Three splitting postures of microservice

    Time:2020-3-30

    We know that microservice is an idea, without definite definition and boundary, such as design principle, which belongs to abstract concept. When the definition is not clear, talking about division is also a kind of talking about each other. Specific problems need specific analysis. Therefore, the division mentioned in this article is not an absolute […]

  • NLP tutorial: how to automatically generate couplets

    Time:2020-3-20

    Demo: https://www.flyai.com/couplets The most important feature of the recurrent neural network is that it can take the sequence as input and output, and the up and down links of the couplet are typical sequence words. So, can we use the neural network to carry out the couplet? The answer is yes. In this project, the […]

  • Why feature normalization / standardization?

    Time:2020-3-18

    Catalog Written in front Common feature scaling methods Comparative analysis of calculation methods Feature scaling required or not When do I need feature scaling? When is feature scaling not needed? Summary Reference resources Blog: blog.shinelee.me | blog Park | CSDN Written in front Feature scaling, commonly referred to as “feature normalization” and “standardization”, is an […]