神经网络

输入 计算 输出

分类问题、michael neisen andrew ng

逻辑回归

支持向量机

激活器  

前向传播

感知器MLP

预测精准度取决于他的权重和偏差

 

 

 

[展开全文]

RELU

Gating units - LSTM, GRU

[展开全文]

Text analytics

NLP - natural language processing

Vector

Dimensionlity reduction

word2vec/GloVe

[展开全文]

Vector algebra

Parallel processing: share mem/distributed computing

FPGA

ASIC (TPU/nervana)

data/model/pipeline parallelism

[展开全文]

Error

TP,FN,FP,TN: confusion marix

recall: TP/sum of positive

precision:TP/data points classified as positive (TP+FP)

multi class classification

[展开全文]

Tensorflow: python lib

Computational graph

Tensorflow vs Theano

Data parallelism

OpenCL support

TensorBoard

[展开全文]

Underfitting & Overfitting

traing set/test set/cross-validation set

[展开全文]

Caffe: compute vision & forecast: C++ lib

with Matlab & python interface

AlexNet & GoogleNet: 2 popular net in Caffe

Blob

[展开全文]

Deeplearning4J: DL4J: java lib

Iterative map-reduce

[展开全文]

Deeplearning4j, Torch, Caffe

Theano for education

[展开全文]

Dato Graphlab Create

GraphNet: CNN for image

MLP (Multilayer perceptron) for others

[展开全文]

Sentiment analysis

MRI, CT scan

[展开全文]

Recursive Neural Tensor Networks (RNTN)

[展开全文]

autoencoder - RBM

backpropagation - loss

use autoencoder to reduce the dimensionality of data

autoencoder is better than PCA

[展开全文]

相关课程

开课日期:开始
免费课 12632
开课日期:开始
机器学习 7452

授课教师

教务组
微信扫码分享课程