- 博客(165)
- 资源 (5)
- 收藏
- 关注
RSS订阅原创 [tableau] sql
“create calculated fields” —>RAWSQL_STR("SELECT distinct dayflagFROM tableau_jp_day.t_s_report_sumpop_timewhere day=%1", [day_parameter])
2018-11-02 18:07:38
443
转载 [sql]加速
# acceleratehttps://www.cnblogs.com/otsf/p/8676487.htmlhttps://blog.csdn.net/kexijun20014/article/details/4654013https://blog.csdn.net/wjq008/article/details/48947223速度：1<2<3<41. &l...
2018-10-01 09:39:34
155
原创 数据分析相关网站
/*留个笔记，应该会很乱。*/相关分析百度指数http://index.baidu.com/#/数据分析网https://www.afenxi.com人人都是产品经理http://www.woshipm.comデータ分析基礎知識https://www.albert2005.co.jp/knowledge/可视化统计学以帮助加深理解...
2018-09-19 14:03:31
242
原创 思维导图
1. Coggle网页版。颜色很好看。问题是没有办法收缩，整个树都是展开状态，看起来会很乱。 2. Processon网页版。mac的safari操作起来经常重启，而且拖动一个branch动不动就卡住了，或者说一直随着光标动来动去的，停不下来。 3. FreeMindsoftware. 并没有安装，不评价。 4. 百度脑图。网页版...
2018-09-14 14:41:08
258
原创 [linux]mailx(gmail,office365)
Gmail参考：https://www.systutorials.com/1411/sending-email-from-mailx-command-in-linux-using-gmails-smtp/https://coderwall.com/p/ez1x2w/send-mail-like-a-bossStep 1:(In fact we have files in "/etc/...
2018-07-12 08:39:34
1982
转载 keras中间层输出
https://keras.io/getting-started/faq/#how-can-i-obtain-the-output-of-an-intermediateOne simple way is to create a new Model that will output the layers that you are interested in:```pythonfrom ker...
2018-03-07 17:31:12
2911
2
原创 [torch] some bugs
error11.”bad argument #1 to ‘set’ (expecting number or torch.DoubleTensor or torch.DoubleStorage at /tmp/luarocks_torch-scm-1-9965/torch7/generic/Tensor.c:1153)”mulLSTMmodel = nn.MapTable(nn.convLSTM(o
2017-12-08 18:53:55
639
原创 [python] read file
read a txt file and remove \n and punctuationsdes_path="event_description/"for i in range(21,41): des_txt = "%s/E0%d.txt" %(des_path, i) with open(des_txt, 'r') as f: d
2017-12-02 15:00:09
491
原创 [tsubame]
install using pippython install --user xxxssd2 TB SSD is installed as local scratch area in each compute node4.4.3.Local scratch area Each node has SSD as local scratch disk space available to your job
2017-12-01 09:31:36
311
原创 [tsubame]python[AssertionError: Cannot find installation of real FFmpeg]
job.sh#bin/bash#qsub -g tga-shinoda ./1_genShot.shinfile=/gs/hs0/tga-xx/xx/annotationFiles/Train-6988-shuffleout_dir=/gs/hs0/tga-xx/xx/videoshot_iccv/shotFramesFFMPEG_PATH=/gs/hs0/tga-xx-deep/xx/libr
2017-11-24 20:13:03
2255
1
原创 [tsubame]install shotdetect
https://github.com/johmathe/Shotdetectlocaldepency for the shotdetection tool: sudo apt-get install libavdevice-dev libavformat-dev libgd2-noxpm(can’t find) libgd2-noxpm-devsudo apt-get install libw
2017-11-22 17:26:22
537
转载 [torch]maskzero-brnn
https://www.bountysource.com/teams/element-research/issues?tracker_ids=13818311require 'nn'require 'rnn'inputdim = 5outputdim = 3seq = 4batch = 2--seqlen x batchsize x inputsizemodel1 = nn.SeqBRN
2017-11-10 13:35:25
296
原创 [torch]creat a new layer
MyMean.luarequire 'nn'require 'os'local MyMean, parent = torch.class('nn.MyMean', 'nn.Sum')function MyMean:__init(maskZero) --input: batch x rdim x cdim --output: batch x cdim
2017-11-06 19:47:39
348
转载 [python][lua]在python里调用Lua
https://stackoverflow.com/questions/30841738/run-lua-script-from-pythonYou can use a subprocess to run your Lua script and provide the function with it’s arguments.import subprocessresult = subprocess.
2017-10-27 13:33:03
1618
1
原创 [paper]summary(event detection)
ACMMMACMMM2015_EventNet: A Large Scale Structured Concept Library for Complex Event从”WIKIHOW”里选出了500个event(共95,321 videos, 4490 concepts). 然后用这些concepts和events建了一个CONCEPT LIBRARY, 就是个树状的.ACMMM2015_Mul
2017-10-06 09:43:32
1897
原创 [tsubame]submit jobs
#!/bin/sh#$ -cwd#$ -l f_node=1#$ -l h_rt=1:00:00#$ -N serial. /etc/profile.d/modules.shmodule load cuda. /gs/hs0/tga-shinoda/15M54105/distro/install/bin/torch-activateth test.lua
2017-09-27 16:47:08
399
原创 [latex]texshop
referenceWe have test.tex, tacl.bib, acl2012.bsttest.tex\begin{document}\maketitle\begin{abstract}%write the abstract here.\end{abstract}\section{Introduction}%write some thing hereYou can cite
2017-08-23 19:28:57
1968
原创 [torch]同时更新多个seqlstm
maptablerequire 'nn'require 'rnn'require 'os'local batch_size = 5local feat_dim = 6local hidden_size = 4local seq_len = 10local num = 10 local lr = 0.01--------initialize model--[[local model
2017-08-23 16:05:35
643
原创 [torch]nngraph parameters of the node
https://github.com/torch/nngraph/issues/131require 'nn'require 'nngraph'require 'rnn'function make_net() local mlp = nn.Sequential(); -- make a multi-layer perceptron mlp:add(nn.Mask
2017-08-07 17:40:07
583
原创 [torch]memory leakage
https://github.com/torch/torch7/issues/229 https://groups.google.com/forum/#!topic/torch7/POColSrmlzk在学校的超级计算机上训练模型,跑了2h15min左右之后说是memory超过了25g, 任务被终止了.于是我在lua里写了print("memory:"..collectgarbage("count
2017-05-25 12:48:23
817
原创 [torch] remove layers from the model
https://github.com/torch/nn/blob/master/doc/containers.mdodel = nn.Sequential()model:add(nn.Linear(10, 20))model:add(nn.Linear(20, 20))model:add(nn.Linear(20, 30))model:remove(2)> modelnn.Sequent
2017-05-22 12:40:18
830
转载 [torch]optim.sgd学习参数
https://stats.stackexchange.com/questions/29130/difference-between-neural-net-weight-decay-and-learning-ratelearning rateThe learning rate is a parameter that determines how much an updating step influ
2017-05-19 15:46:40
9599
原创 [torch]Getting the Output of a Layer
https://groups.google.com/forum/#!topic/torch7/R9DAdx95aqcrequire 'cutorch'require 'cunn'require 'rnn'require 'os'--[[net = nn.Sequential()net:add(nn.FastLSTM(10,3))net:add(nn.Linear(3,4))net:a
2017-05-10 18:17:49
448
原创 [torch]create a new criterion(cross entropy)
http://blog.csdn.net/u010167269/article/details/51966427IntroductionCrossentropyCriterion in torchCrossEntropyCriterion and ClassNLLCriterion only support with hard targets.For examples, i
2017-05-08 19:20:47
3805
原创 [torch]criterion/maskzero
require 'nn'require 'rnn'maxzero = 1if maxzero == 1 then y_ = torch.Tensor{{1,5,2},{0,5,2},{0,0,0}} y = torch.Tensor{{2,3,1},{0,5,4},{0,0,0}} y = torch.Tensor{{2,3,1},{0,5,4},
2017-05-06 13:15:14
736
原创 [总结]神经网络・压缩 compression(cnn,rnn)
distillation:papers:NIPS2014_Distilling the Knowledge in a Neural Network_Hiton FitNets: Hints for Thin Deep Nets Dropout Distillation Distilling Knowledge to Specialist Networks for Clustered Class
2017-04-17 12:40:27
2109
转载 [torch]parameters(clone/copy/initialize)
https://github.com/torch/nn/blob/master/doc/module.md-- make an mlpmlp1=nn.Sequential();mlp1:add(nn.Linear(100,10));-- make a copy that shares the weights and biasesmlp2=mlp1:clone('weight','bias');
2017-03-24 15:14:23
914
转载 [torch] rnn training
http://christopher5106.github.io/deep/learning/2016/07/14/element-research-torch-rnn-tutorial.html https://github.com/torch/optim/blob/master/doc/intro.md http://rnduja.github.io/2015/10/26/deep_lear
2017-03-24 14:17:36
311
原创 attention model
How did I select papers?First, I tried to search for “attention” in CVPR2014-2016, ICCV2009-2015 and ACMMM2012-2015. However, there are only a few papers containing this key words. Then, I searched f
2017-03-01 16:16:02
1309
原创 [python]unicode,utf8
I’m reading a json file by python.[ { "id": 1, "qas": [ { "image_id": 1, "qa_id": 10000000, "answer": "グレー", "a_objects": [], "question": "右側の男性は
2017-02-15 20:15:29
471
原创 [torch]Save initial state(fastlstm)
beforelocal FastLSTM, parent = torch.class("nn.FastLSTM", "nn.LSTM")-- set this to true to have it use nngraph instead of nn-- setting this to true can make your next FastLSTM significantly fasterFas
2017-02-15 15:46:28
647
原创 [torch]nn内部函数?
https://bigaidream.gitbooks.io/subsets_ml_cookbook/content/dl/lua/lua_module.html[output] forward(input)Takes an input object, and computes the corresponding output of the module.After a forward(), the
2017-02-08 14:09:44
1071
原创 [python]画图
generalimport sysimport numpy as np import matplotlibmatplotlib.use('Agg') from matplotlib.pyplot import plot,savefigimport matplotlib.pyplot as pltdef isfloat(value): try: float(value)
2016-11-21 10:12:55
553
原创 [linux]terminal
Are there any methods to run experiments if when I am absent? Now I use putty to connect to the server and run the code, however, once I leave the computer more than 10 minutes, it will be disconne
2016-10-25 12:49:27
468
转载 [rnn]BPTT_梯度消失/爆炸问题
http://www.wildml.com/2015/10/recurrent-neural-networks-tutorial-part-3-backpropagation-through-time-and-vanishing-gradients/翻译: https://zhuanlan.zhihu.com/p/22338087随时间的反向传播（BPTT）让我们先迅速回忆一下RNN的基本公式，注
2016-10-18 14:20:16
13208
转载 [python]image(reshape,overlay)
import sysimport numpy as np#import scipy.miscfrom scipy import ndimage, miscfrom PIL import Imageimgpath="demo_img1.jpg"#1.generate Attention Map#files=["word.txt","phrase.txt","question.txt"]
2016-10-13 16:58:27
2510
转载 [http]get,post
web API: https://labs.goo.ne.jp/api/jp/morphological-analysis/import requests r = requests.post(‘https://labs.goo.ne.jp/api/morph‘, json={“app_id”:”4a875765817519b8e0ced23710a146d60f617e5b0d5a44e7109
2016-09-28 09:31:49
316
原创 [torch]read file
read file(torch)feattemp=torch.Tensor(batchSize,seqLengthMax,featdim)-- ...file = io.open("feat.txt", "r")io.input(file)temp=io.read()for m=1,featdim do print("m"..m) feattemp[i][j][m]=temp:
2016-07-29 19:27:20
1310
matlab画图不要白边
2015-04-21
matlab_hog
2015-01-21
matlab提取hsv，rgb，lab
2014-11-07
KLT_MATLAB
2013-10-15
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人