Tag：accuracy
-
Time：2021-7-27
In this paperParameter adjustment record 25On the basis of, the number of neurons in the adaptive parameterized relu middle layer is increased from 2 to 4, and a dropout layer is added to continue to test its effect on the cifar10 data set. Basic principle of adaptive parameterized relu: Keras program: #!/usr/bin/env python3 # -*- […]
-
Time：2021-7-20
Take the kernel version 3.10 as an example, the kernel version 4.1 + has some changes when dealing with fin-wait-2, which will be mentioned later The code should be reasonably simplified TL；DR Time of Linux TCP_ The timeout of wait status is 60 seconds by default and cannot be modified Fin of Linux TCP_ WAIT_ […]
-
Time：2021-7-8
This paper proposes a neural network architecture search method for mobile terminal. There are two main ideas in this method. Firstly, the multi-objective optimization method is used to integrate the time-consuming of the model on the actual device into the search, and then the hierarchical search space is used to keep the diversity of the […]
-
Time：2021-6-24
PS: there is a problem of precision loss in the precision processing of the project today. Let’s share the solution of the expansion package with you. Note: this problem is possible, not inevitable ① Install expansion pack go get github.com/shopspring/decimal ② How to use it //Set the number of reserved digits first Decimal. Divisionprecision = […]
-
Time：2021-6-8
The idea of netadapt is ingenious and effective. It divides the optimization target into several small targets, and introduces the actual indicators into the optimization process. It can automatically generate a series of platform related simplified networks. It not only has fast search speed, but also has good performance in accuracy and delay Source: Xiaofei’s […]
-
Time：2021-6-7
Fbnet series is a lightweight network series completely based on NAS method. The shortcomings of current search methods are analyzed and innovative improvements are gradually added. Fbnet combines DNAs and resource constraints, fbnetv2 adds channel and input resolution search, and fbnetv3 uses accuracy prediction for fast network structure search Source: Xiaofei’s algorithm Engineering Notes official […]
-
Time：2021-5-6
Author muskan097Compile VKSource: analytics vidhya brief introduction You have successfully built the classification model. What do you do now? How do you evaluate the performance of the model, that is, the performance of the model in terms of prediction results. To answer these questions, let’s look at the metrics used in evaluating classification models through […]
-
Time：2021-4-17
In the nasnet search space, compared with reinforcement learning and random search, the algorithm is simple enough, and can search higher quality models faster. Amoebanet-a can achieve SOTA on Imagenet Source: Xiao Fei’s algorithm Engineering Notes] official account Paper: regulated evolution for image classifier architecture search Address: https://arxiv.org/abs/1802.01548 Introduction Neural network structure search network has […]
-
Time：2021-3-26
data set First, choose cifar-10 for the dataset. This data set containsTen categoriesPictures of each category6000Zhang32 x 32The total number of pictures60000Two pictures, of which50000Training pictures,10000This is a test picture. It’s downloaded herepythonCorresponding version: read file stayCifar-10We can find the instance code and read it according to our own file directory CIRFA_DIR = “../data/cifar-10-batches-py” def […]
-
Time：2021-3-26
Scientific counting is a method of counting.Features: precision loss, space saving What is scientific counting Express a number as a and aThe form of multiplication of n-th power of 10(1 ≤| a | < 10, a is not a fraction, n is an integer).19971400000000 == 1.99714×10^13 == 1.99714e13 When we want toTo mark or calculate […]
-
Time：2021-3-20
By behic guvenCompile VKSource: towards Data Science In this article, I will introduce a machine learning method called supervised learning. I’ll show you how to use scikit learn to build KNN classifier model. This will be a practical exercise, and we will be able to learn while practicing knowledge. As our classifier model, we will […]
-
Time：2021-3-12
Refinedet can be regarded as the combination of SSD, RPN and FPN algorithms. Its main idea is: fast RCNN and other two stage algorithms regress the box twice, so the precision is high, but the speed is slow; Yolo and other one stage algorithms regress the box only once, which is fast, but the precision […]
