Wu Enda deep learning L1 assignment 2 -- implementation of classification neural network with single hidden layer
Implementation of 2-classification neural network with a single hidden layer Use neurons with nonlinear activation functions, such as tanh Some notes do homework Import commonly used packages and toolsets provided by teachers import numpy as np import sklearn import matplotlib.pyplot as plt frUTF-8...
Posted by Snart on Sun, 10 Oct 2021 19:23:08 +0530
Circular Neural Network from Zero
import math import torch from torch import nn from torch.nn import functional as F from d2l import torch as d2l batch_size, num_steps = 32, 35 #bach_size is the number of samples in batch and num_step is the sequence length of a sample train_iter, vocab = d2l.load_data_time_machine(batch_size, UTF-8...
Posted by GBahle on Sun, 10 Oct 2021 21:46:50 +0530
YOLOv4 based on source code detail analysis (positive and negative sample allocation)
Difference between v3 and v4 in positive sample selection: I have reservations about this, because bloggers read the source code and found that YOLOv4 is still the one with the largest IOU in the gt box for each gt box, and then assigned to the layer with the largest IOU to predict. The locatioUTF-8...
Posted by timbr8ks on Mon, 11 Oct 2021 23:42:14 +0530
Image geometric transformation -- zoom, translation, mirror, rotation
1, Fundamentals of image geometric transformation 1. Geometric transformation of image refers to the change of size, shape and position of the original image as required. 2. The pixel value of the image is not changed, but the geometric position of the pixel is changed. 3. For the original imaUTF-8...
Posted by tadapapaya on Tue, 12 Oct 2021 01:55:13 +0530
Chapter 6 Part-3 of the book brushing notes of Python deep learning uses the Embedding layer to learn word Embedding
Use word Embedding layer (Embedding layer) There is also a powerful method for associating words with vectors, which is to use a more dense word vector, also known as word embedding. Compared with one hot coding, it is as follows: densitydimensionGeneration methodOne-hotsparseHigh dimensionHarUTF-8...
Posted by mbeals on Tue, 12 Oct 2021 05:08:18 +0530
Learn AI from Li Mu -- anchor box code analysis -- 1
Learn AI from Li Mu - anchor box code analysis - 1 Introduction to anchor frame Target detection algorithms usually sample a large number of regions in the input image, then judge whether the region contains the target of interest, and adjust the region edge, so as to more accurately predict thUTF-8...
Posted by the_manic_mouse on Tue, 12 Oct 2021 12:15:16 +0530
PyTorch visualizes and dynamically adjusts the learning rate lr_scheduler
Stick to blogging and share your gains in study and work Make a note of yourselfRecord and summarize knowledge points to deepen understandingGive some help to people in need, step less on a pit and take more steps Try to arrange the layout in an appropriate way, with both graphics and text If yUTF-8...
Posted by geoffism on Tue, 12 Oct 2021 13:13:26 +0530
Python OpenCV implements Log Gabor filter (extended by LGHD descriptor)
introduction The author came into contact with many descriptors when studying the registration of infrared image and visible image. Among them, the Log Gabor filter of LGHD descriptor is very interesting to share with you LGHD(Log-Gabor Histogram Descriptor) The idea of descriptor is that the dUTF-8...
Posted by gtanzer on Wed, 13 Oct 2021 19:18:42 +0530
DeepFM detailed process analysis and modeling demonstration
Process Analysis There are many blogs or papers explaining the DeepFM theory, which will not be repeated here. Here is how the model is used and what the algorithm does to the data throughout the process. First of all, this picture: At a glance, is it super complex? Yes, but don't panic. Let's UTF-8...
Posted by elfyn on Wed, 13 Oct 2021 21:40:18 +0530
Brief reading and reproduction of ST-GCN papers
code: https://github.com/yysijie/st-gcn Ordinary convolution and graph convolution Ordinary convolution Before understanding graph convolution, we need to understand the traditional convolution. For a 3 × For the convolution operation of 3, it is equivalent to 3 × The convolution kernel of 3 sUTF-8...
Posted by Dimensional on Thu, 14 Oct 2021 06:00:18 +0530