Building simple features and understand the underlying patterns that can be applied to real-world examples
Create in 2022.02.14. Monday; Finish Feature 1 of Netflix 1. Netflix Features Netflix is the biggest video streaming platform in the world, offering movies, seasons, biographies, reality shows, and more.
Their video repository is growing significantly. So the engineering team at Netflix keeps trying to find better ways to display content to their consumers.
1. Weight Initialization A well chosen initialization can:
Speed up the convergence of gradient descent Increase the odds of gradient descent converging to a lower training (and generalization) error Gradient Exploding and Vanishing Training a neural network requires specifying an initial value of the weights. A well chosen initialization method will help learning.
if W > I (Identity Matrix): in DNN, the activations can explode if W < I: the activations will decrease exponentially and then vanish Zero Initialization：
Regularize Deep Neural Networks 这篇笔记记录了深度神经网络的正则化，在【统计学习笔记 03 世上没有解千愁的酒】中，我记录了“没有任何一种模型或算法可以在各种数据集中完胜其它所有的算法，解决所有的问题”，当我们把神经网络的层数加深，神经元节点变多，使其能“深度学习”更多的更复杂的更灵活的非线性函数的同时，也使得模型产生了**过拟合(overfit)**问题 —— 酗酒(overdrink)看似可以完全地麻痹自己让自己完全忘记了所有问题、幻想自己活在了一个完美的温柔乡(low training set error)，但酒醒之后却仍然无力解决任何现实的问题，仍然脆弱至极连站都站不起来更不要说能跑起来(high testing set error)。我们需要自律，深度神经网络亦如是 —— 正则化(Regularization)的加入使得神经网络避免陷入过度拟合的泥沼，在这篇笔记提到的四种正则化的方法(L2, Dropout, Data Augment, Early Stop)可以让其在测试集中表现的仍然良好。
1. Train/Dev/Test Sets Applying Machine Learning is highly iterative process
Idea ==> Code ==> Experiment We have to go through the loop many times to figure out your hyperparameters. We build a model upon training set then optimize hyperparameters on dev set as much as possible.
Step 1: Search for IntelRealSense github
https://github.com/IntelRealSense/librealsense Step 2: Click the tag => release => latest
https://github.com/IntelRealSense/librealsense/releases Click the latest tag (2.51.1 on Oct 9, 2022) and install the SDK on linux https://github.com/IntelRealSense/librealsense/blob/master/doc/distribution_linux.md Step 3: Register the server’s public key and add server to the list of repositories
Step 4: Install the libraries and the developer and debug packages
sudo apt-get install librealsense2-dkms sudo apt-get install librealsense2-utils sudo apt-get install librealsense2-dev sudo apt-get install librealsense2-dbg Step 5: Reconnect the Intel RealSense Depth Camera (D435i) and run:
conda install -c anaconda h5py conda install -c conda-forge opencv git clone https://github.com/qqwweee/keras-yolo3.git conda install -c conda-forge tensorflow conda install -c conda-forge keras conda install -c anaconda pillow conda install -c conda-forge matplotlib import tensorflow.python.keras.backend as K import pyrealsense2.pyrealsense2 as rs
0. Check System, CPU, GPU 1 2 $ sudo lshw -C display $ lscpu GPU: NVIDIA GeForce GTX 1060 3GB CPU: Intel i7-5820K Download and use Miniconda (latest Version 3.9) https://docs.conda.io/en/latest/miniconda.html#linux-installers Conda is an open source package management system and environment management system that runs on Windows, macOS and Linux. Conda quickly installs, runs and updates packages and their dependencies.
