Please find instructions for this assignment in the attached document. You need to complete all...

Question

Please find instructions for this assignment in the attached document. You need to complete all questions for short answer please provide your answers in sentences.

1. (20%) What about the main inventions for GoogLeNet, ResNet, SENet, and Xception? 2. (25%) Can you compare the difference between the word-level embedding and character-level embedding from text classification perspective? 3. (25%) BERT, proposed by Google, is another popular word embedding technique. Can you explain this and compare this with Word2Vec (Skip-gram, CBOW)? 4. (15%, coding assignment) Try to follow the code about style transfer (https://www.tensorflow.org/tutorials/generative/style_transfer) with DL, you can select any style image to transform content image of your wonderful face (Your face photo) into your DL face. For this assignment, beside turn in your implementation code. You also have to submit TA your DL face image (try to adjust your DL face size to fit US passport photo size) with your name. We will have a Beauty Pageant for our class based on your DL face if time enough! 5. (15%, coding assignment) Implement Word2Vec from scratch, you can reference this website (https://towardsdatascience.com/an-implementation-guide-to-word2vec-using-numpy-and-google-sheets-13445eebd281). You can choose your own texts corpse for training Word2Vec embedding model. Then, apply your own word-embedding model to perform sentiment classification for IMDB dataset (https://keras.io/api/datasets/imdb/) with CNN architecture. Try to demonstrate the effects of hyperparameters, e.g., parameters about Word2Vec and CNN model, to the classification performance.

cnn-wowfmv54.docx

Vicky · Accepted Answer

1. (20%) What about the main inventions for GoogLeNet, ResNet, SENet, and Xception?
Ans. GoogLeNet: Also known as GoogLeNet , it is a 22-layer network that won the 2014 ILSVRC Championship.
a) The original intention of the design is to expand the width and depth on its basis .
b) which is designed motives derived from improving the performance of the depth of the network generally can increase the size of the network and increase the size of the data set to increase, but at the same time cause the network parameters and easily fit through excessive , computing resources inefficient and The production of high-quality data sets is an expensive issue.
c) Its design philosophy is to change the full connection to a sparse architecture and try to change it to a sparse architecture inside the convolution.
d) The main idea is to design an inception module and increase the depth and width of the network by continuously copying these inception modules , but GooLeNet mainly extends these inception modules in depth.
There are four parallel channels in each inception module , and concat is performed at the end of the channel .
1x1 conv is mainly used to reduce the dimensions in the article to avoid calculation bottlenecks. It also adds additional softmax loss to some branches of the previous network layer to avoid the problem of gradient disappearance.
ResNet: ResNet is a network structure proposed by the He Kaiming, Sun Jian and others of Microsoft Research Asia in 2015, and won the first place in the ILSVRC-2015 classification task. At the same time, it won the first place in ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation tasks. It was a sensation at the time.
ResNet, also known as residual neural network, refers to the idea of ​​adding residual learning to the traditional convolutional neural network,

1. (20%) What about the main inventions for GoogLeNet, ResNet, SENet, and Xception? 2. (25%) Can you compare the difference between the word-level embedding and character-level embedding from text...

Answer To: 1. (20%) What about the main inventions for GoogLeNet, ResNet, SENet, and Xception? 2. (25%) Can you...

Answer To This Question Is Available To Download

Related Questions & Answers

Submit New Assignment