cslt Wiki - 用户贡献 [zh-cn]

2019-02-20

2019-02-20T03:43:30Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
*
||
*
||
*
|-

|-
|Jiayao Wu
||
*finished several experiments on node-sparseness
||
*keep on doing experiments
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*1,use the ASR test_data set ref.txt to test the original bert and then compare with the fine-tuned Bert.
*2,compare the difference between hyp.text which comes from ASR test results, and investigate how to mask the error in the sentence of the ASR result.
*3,set a rule among all the result(hyp.text) to find out the error to be masked.
||
*
||
*
|-

|-
|Dan He
||
*
||
*Continue to study the time complexity of TT-decomposition
||
*
|-

|-
|Yang Zhang
||
*
||
*
||
*
|-

|-
|Wenwei Dong
||
*
||
*
||
*
|-

|}

2019-02-04

2019-02-16T12:12:00Z

Hedan：

$ Happy new year :) $

{| class="wikitable"
!People !! Day home !! Day back !! Life Tracking
|-

|-
|Yibo Liu
||
1月31日
||
2月13日
||

|-

|-
|Xiuqi Jiang
||
1月23日
||
2月24日
||

|-

|-
|Jiayao Wu
||
1月25日
||
2月13日
||

|-

|-
|Zhaodi Qi
||
1月31日
||
预计2月13日（抢票中）
||

|-

|-
|Jiawei Yu
||
1月30号
||
预计2月11号（不过还没抢到票）
||

|-

|-
|Yunqi Cai
||

||

||

|-

|-
|Dan He
||
1月25日
||
2月21日
||

|-

|-
|Yang Zhang
||
01-24
||
02-17
||

|-

|-
|Wenwei Dong
||
1月30号
||
预计2月14号
||

|-

|}

2019-01-23

2019-01-23T04:32:53Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
* Started to reconstruct the vivi code with better structures.
||
* Continue the code reconstruction. Especially need to build proper models for planning and post processing.
||
*
|-

|-
|Xiuqi Jiang
||
* Designed a better code structure for further experiments.
* Improved vivi2.0 and made some adjustments to .sh script.
||
* Build codes under the new structure.
||
*
|-

|-
|Jiayao Wu
||
* do experiments on node_sparseness and update it on cvss
* re-label some data
||
* keep on doing experiments on pruning
* get familiar with pytorch
||
*
|-

|-
|Zhaodi Qi
||
* Reduce the lid model and test the results
* Test test set of different channels
* Wrote a model based on asr(tdnn-f)-lid(tdnn) (similar to PTN) to solve channel inconsistency
||
* Complete the asr-lid model
||
*
|-

|-
|Jiawei Yu
||
* wrote a tensorflow learning document, and I have not completed it.
* read some papers about attention and find some attention code in GitHub.
||
* try to run these attention code figure out mechanism of this code.
||
*
|-

|-
|Yunqi Cai
||
*1,figured out how the Bert model create the pretraining data and do the pretraining.
*2,try to use the Bert to do the error correction of a text sentence.
*3,re-label some ASR data
*4,Test vivi2.0 model
||
*Construct a Text sentence error correction model
||
*
|-

|-
|Dan He
||
*Do experiments on comparing test time and update it on cvss
*Read the experiment code carefully
||
*Directly decompose the trained parameters and put them into the network for retraining.
||
*
|-

|-
|Yang Zhang
||
* 1. remodified nginx configuration and changed the server networking structure
* 2. tried to learn vae and did a [https://github.com/hwalsuklee/tensorflow-mnist-VAE test] in wolf server
||
* continue to learn and test VAE
||
*
|-

|}

2019-01-23

2019-01-23T03:17:48Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Designed a better code structure for further experiments.
* Improved vivi2.0 and made some adjustments to .sh script.
||
* Build codes under the new structure.
||
*
|-

|-
|Jiayao Wu
||
* do experiments on node_sparseness and update it on cvss
* re-label some data
||
* keep on doing experiments on pruning
* get familiar with pytorch
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*
||
*
||
*
|-

|-
|Dan He
||
*Do experiments on comparing test time and update it on cvss

||
*
||
*
|-

|-
|Yang Zhang
||
*
||
*
||
*
|-

|}

2019-01-16

2019-01-16T04:21:21Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Focused back on the quatrain generation, thinking of the drawbacks of current model.
* Tried to weaken the attention mechanism between sentences.
||
* Add VAE into model and try to generate instead of predicting.
||
*
|-

|-
|Jiayao Wu
||
*
||
*
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*1.Run through the bert model. 2.study the details of the bert model.
||
*
||
*
|-

|-
|Dan He
||
*Doing some comparative experiments about TT-decompositions.
||
*Summarize the results of the comparative experiments and initially complete the relevant research on TT-decomposition.
||
*
|-

|-
|Yang Zhang
||
* exam week.
||
* (I will finish all my exams on Friday night)
||
*
|-

|}

2019-01-09

2019-01-09T02:33:56Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
*
||
*
||
*
|-

|-
|Jiayao Wu
||
* finished the speech book -- [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/c/c5/Kws.pdf kws]
* doing experiment of node -pruning on WSJ chain model
||
* keep on node-pruning research
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*
||
*
||
*
|-

|-
|Dan He
||
* Compared the inference time, but the results is not well.
*After TT decomposing the two full-connection layers, it is found that the testing accuracy is very low.
||
*According to some problems found in the experiment, continue to do comparative experiments and analyze the reasons.
||
*
|-

|-
|Yang Zhang
||
* Wrote a brief [https://github.com/zyzisyz/VPR-wx-client document]
* submit the source code to [https://gitlab.com/zyzisyz/nebula-listen GitLab].
||
* revise for my school subjects and prepare for the final-term examinations.
||
*
|-

|}

文件:ML学习笔记何丹.pdf

2019-01-07T00:55:23Z

Hedan：Hedan上传“文件:ML学习笔记何丹.pdf”的新版本

ML book

2019-01-04T07:45:03Z

Hedan：

'''Introduction to Modern Machine Learning Technology, by Dong Wang, [http://mlbook.cslt.org Official website].
'''

----

Reading notes of interns
*[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/a/a6/CSLTBOOK读书笔记.pdf Jiayao Wu (吴嘉瑶)]
*[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/e/ec/读书笔记——于嘉威.pdf Jiawei Yu (于嘉威)]
*[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/b/b1/机器学习读书笔记_姜修齐.pdf Xiuqi Jiang (姜修齐)]
*[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/a/ac/ML学习笔记_何丹.pdf Dan He (何丹)]

文件:ML学习笔记何丹.pdf

2019-01-04T07:33:43Z

Hedan：

2019-01-02

2019-01-03T03:08:57Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Made further adjustments to code, deleting unnecessary files and uploading generated samples to dir 'predict/results'.
* More models have been trained and compared.
* Collated notes on ML book and uploaded them on wiki.
||
* Try to train a model generating length-unfixed texts.
||
*
|-

|-
|Jiayao Wu
||
*
||
*
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*
||
*
||
*
|-

|-
|Dan He
||
*Compared the size of a fully connected layer after tt decomposition through experiment, at the same time compared the train loss, validation loss and test accuracy between the two case.
||
*Further compared the inference speed, the precision after retrain.
*Try to decompose more layers of fully connected layers.
||
*
|-

|-
|Yang Zhang
||
* Released and published our WeChat app(星云听).
* cleaned my code
* wrote document for this project (have not finished yet).
||
* continue to write document.
||
*
|-

|}

文件:TT-Loss.png

2019-01-02T11:45:27Z

Hedan：

2018-12-26

2018-12-26T03:01:07Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Finished cleaning up the code of ViVi and completed the document.
* Did some reconstruction and made training and predicting processes more clearly.
||
* Try to build demo of generating iambics.
||
*
|-

|-
|Jiayao Wu
||
*
||
*
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
* run thch30 recipe for d-vector and asr model
* read max-margin code and paper
||
* extract d-vector and phone vector by max-margin
||
*
|-

|-
|Yunqi Cai
||
*Learn how to use Xhell, Linux and Kaldi.
*Be familiar with the functions of Kaldi through run "Thchs30" on it.
||
*Run through "Thchs30" on Kaldi and learn every steps
||
*
|-

|-
|Dan He
||
*Run the experimental code of the paper about tensorizing nerual networks
*Have not finished yet
||
*Contiune running the code
*Understand the experimental code
||
*
|-

|-
|Yang Zhang
||
* Built a more reliable speaker recognition service in our server.
* Now, we can test and use 3 kind of VPR models(ours, Microsoft and SpeakIN) in our Wechat APP.
||
* Fix some small bugs, beautify web UI design.
* Prepare for my final-term examinations.
||
*
|-

|}

2018-12-19

2018-12-19T03:12:03Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Reconstructed codes and made adjustments on training part.
||
* Focus on generating Song Iambics.
||
*
|-

|-
|Jiayao Wu
||
*
||
*
||
*
|-

|-
|Zhaodi Qi
||
* 1. Completed 1th and 2th steps of the DID system.
* 1th : Data preparation (train dev test (1/5 data selected from dev) wav.lst & word.txt) and Dictionary preparation, dict.txt
* 2th : Generate batch data and feed it to the next acoustic model framework
||
* 1. Reference paper completes acoustic model of dialect
* paper : 2015， Baiguang Shi et.al, An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*
||
*
||
*
|-

|-
|Dan He
||
*Run the experimental code of the paper about tensorizing nerual networks
*Have not finished yet
||
*Contiune running the code
*Understand the experimental code
||
*
|-

|-
|Yang Zhang
||
* 1. investigated VPR service and wrote a investigation report.
* 2. Tested Microsoft and SpeakIn's VPR interface.
||
* try to release a stable WeChat APP at the end of this month.
||
*
|-

|}

2018-12-12

2018-12-12T02:30:42Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Learned TensorFlow documents about seq2seq model and made adjustments on vivi, especially the buckets parts
* Had some problems on encoder-memory and decoder-memory, and was trying to figure out the difference
||
* Make adjustments on vivi
* Reproduce codes on Song Iambics generation
||
*
|-

|-
|Jiayao Wu
||
*
||
*
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*
||
*
||
*
|-

|-
|Dan He
||
*Read papers about tensor decompositions
*Prepared a paper sharing about tensorizing neural work
||
*Run the code for the experimental part of the paper
*Learn more about the details
||
*
|-

|-
|Yang Zhang
||
* 1.refactored my code and accelerated VPR speed.
* 2.modified some references error in Machine Learning book.
||
* try to release a stable app as soon as possible.
||
*
|-

|}

2018-12-05

2018-12-05T05:28:27Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
*
||
*
||
*
|-

|-
|Jiayao Wu
||
*
||
*
||
*
|-

|-
|Zhaodi Qi
||
*
||
*
||
*
|-

|-
|Jiawei Yu
||
*
||
*
||
*
|-

|-
|Yunqi Cai
||
*
||
*
||
*
|-

|-
|Dan He
||
*
||
*1.Read the first part of the book "Tensor Networks for Dimensionality Reduction and Large-Scale Optimization" written by Andrzej Cichocki and Ivan Oseledets.
*2. Initial understanding of the CP decomposition and Tucker decomposition,but the TT decomposition has not yet understand.
||
*1.Continue read the book to study the TT decomposition.
*2. Prepare a paper sharing about tensor decomposition.
|-

|-
|Yang Zhang
||
*
||
*
||
*
|-

|}

2018-11-28

2018-11-28T04:30:13Z

Hedan：

{| class="wikitable"
!People !! Last Week !! This Week !! Task Tracking (DeadLine)
|-

|-
|Yibo Liu
||
*
||
*
||
*
|-

|-
|Xiuqi Jiang
||
* Have been following CS224n and learning details about language model.
* Read codes and watched over-fitting problems during practice.
* Compared codes on couplet generating and ViVi.
||
* Read the paper "Generating Sentences from a Continuous Space” and work out how VAE goes for sentences.
* Complete current models.
||
*
|-

|-
|Jiayao Wu
||
*Read the paper about TDNN-F and the corresponding codes in kaldi.
*establish the TDNN-F baseline for WSJ.
||
*There still exists some bugs in kaldi. solve the configuration problems for kaldi.
*run the baseline but not finished yet.
||
*
|-

|-
|Zhaodi Qi
||
* Completed the baseline.
* Learned a little tensorflow.
||
* practice tensorflow and read coded about acoustic model.
* I want to install cuda9, it is convenient to install the following Driver Version >=384.+ for the machine?
||
*
|-

|-
|Jiawei Yu
||
* Read the Lantian's phd graduation thesis
* Learn CNN structure for SRE
* Adapt BNF to SRE
||
* Continue investigate the latest paper on NN model for SRE
* Complete the BNF experiment
||
*
|-

|-
|Dan He
||
*Read some paper about tensorizing neural network
*Try to know that how to reduce the number of parameters by tensor decomposition
||
*Continue learn TT-format to represent the dense weight matrix of the fully-connected.
||
*
|-

|-
|Yang Zhang
||
* review school subjects and prepare for my mid-term exams.
* built a basic web-client in WeChat (with my classmate's help)
* Applicated a domain name and SSL certificate, and wrote some back-end code (后端代码).
* made a HTTPs connection and a web socket connection between WeChat web-client and server.
||
* try to finish this WeChat app as soon as possible.
||
*
|-

|}

cslt Wiki - 用户贡献 [zh-cn]

2019-02-20

2019-02-04

2019-01-23

2019-01-23

2019-01-16

2019-01-09

文件:ML学习笔记 何丹.pdf

ML book

文件:ML学习笔记 何丹.pdf

2019-01-02

文件:TT-Loss.png

2018-12-26

2018-12-19

2018-12-12

2018-12-05

2018-11-28

文件:ML学习笔记何丹.pdf

文件:ML学习笔记何丹.pdf