3060m computing power
The South Station of future science and technology city of Beijing metro line 17 has started construction at the end of 2015, and it is planned to complete the whole line in 2019.
as a subway line running through the north and south, Beijing metro line 17 has 20 planned stations, namely, the north of future science and Technology City, the south of future science and Technology City, Tiantongyuan East, qingheying, warrior camp, wangjingxi (transfer to line 13 and line 15) Taiyanggong (transfer to line 10), Xibahe, xiangheyuan, worker's Stadium (transfer to line 3), Dongdaqiao (transfer to line 6), YONGANLI (transfer to line 1), guangqumenwai (transfer to line 7), Panjiayuan West, Shilihe (transfer to lines 10 and 14), Shibalidian [5], Chaoyang port, beishenshu, ciqubei, Ciqu (transfer to Yizhuang line), " South front area of Yizhuang railway station< br/>
Shanghai rail transit line 17, the former Qingpu section of Metro Line 2, system code R2W. Hereinafter referred to as "line 17", it is an important rail transit line planned and constructed in Shanghai ring the 12th Five Year Plan period. It starts from Hongqiao Railway Station in the East (the station transfers with line 2 on the same platform), runs parallel westward along the south side of Songze Avenue, crosses G15 Shenhai expressway, and then turns along Yinggang East Road and Yinggang road to Qingpu City and Dianshanhu new city in the West [2]. After entering Zhujiajiao area, it runs parallel to Dongfang Lvzhou along the south side of national highway 318, with a total length of 35.30 km and 13 stations (including Hongqiao Railway Station). The construction of the main line has been started in 2013-2014. The total construction period of the line is 48 months. It is planned to be completed and opened to traffic by the end of 2017
line:
Hongqiao Railway Station (interchange to
Metro Line 2 & # 92; 10) (2-storey underground side station, transfer with line 2 on the same platform; Line 10 must pass through the station hall
National Exhibition and convention of China Expo North
Center (n.) (Island Station on the second floor underground, Zhuguang Road, Songze Avenue, reserved transfer line 23)
Panlong Rd. (Island station on the second floor underground, Panlong Road, Songze Avenue)
Xuying Rd.
(Island Station on the third floor elevated, Songze Avenue, Xuying Road, and Xujing depot)
Xujing Beicheng (elevated 2-storey Island Station, Songze Avenue, Xule Road, and Xujing depot entrance and exit line)
Jiasong Rd. (M.) (elevated 3-storey side station, Yinggang East Road, Jiasong Middle Road)
Zhaoxiang (elevated 3-storey side station, Huijin Rd.
Huijin Rd.
Qingpu (Island Station with 2 floors underground, Qingsong road outside Yinggang East Road)
caoying Rd, Yinggang Road, caoying Road)
Dianshanhu Avenue
Ave.
Zhujiajiao (elevated 3-storey Island Station, Huqingping highway, Zhufeng highway and Zhujiajiao parking lot)
Oriental
land, To the south of Huqingping highway, opposite to Dongfang Lvzhou square, West extension conditions are reserved)
operation time:
first and last bus: 6:00-23:00
, initial running interval: 4 minutes
West extension is reserved to connect with Suzhou rail transit line S2B
Analysis on the trend of Cheng Metro Line 10 (possibly corresponding to the original planning line R1): taipingyuan station Bayi Furniture City Jinqiao Hospital Station Airport T1 station airport T2 station Shuangliu west station Xinjin County Cheng Metro Line 11 (without the corresponding original drawing planning, So we can only guess by the starting point position and divergence direction): Liulichang East --- Racing Museum Station --- Sanchahu Cheng Metro Line 12 trend analysis, Cheng Metro Line 13 trend analysis (may correspond to the original planning R2 line): Shunjiang road - Sanguantang - CHUANSHI - yancao company - sanshenghuaxiang - longhuasi Station - economic development Zone South Station - Shiji Avenue Station - dongshanguo Trend analysis of Cheng Metro Line 14 at jixincheng station (possibly corresponding to the original planned R3 line): Cheng East Station - Shiling Forest Park South Station - Xihe Town Station - Xiping Town Station - Luodai Town Station - honghuatang Station - Baiguo town (the terminal location of the line shows a planned airport, so it is estimated that the trend of line 14 will be adjusted according to the site selection of the airport, Anyway, there will be a metro line from the east railway station to the second airport) analysis on the trend of Cheng Metro Line 15 (which may correspond to the original planning line R4): ma'anbei road - Bali community station - Maliu Station - shuangbayan Station - poly community - Hai Road Station - Chuanyin Station - Petroleum University Station - New East Station -... - Qingjiang urban area -... - Analysis on the trend of Cheng metro line 16 in Jintang county (which may correspond to Original planning R5 north branch line: Guanghua Village Station (along Guanghua Avenue) - Fengxiang station Yongquan Station - Fengxi Avenue middle station - Laifeng Road Station - Tashui town line 17 trend analysis (may correspond to the original planning R5 south branch line): Dashi West Road - Qingshui River Bridge - Jitou Town Station - Yongkang Forest Park Station - Jiujiang North Station - Jiaolong Instrial Port Station - Jiaolong Instrial Port South Station - Jinwan community station - North Pengzhen Station - South Pengzhen Station - Banqiao Village original planning R0 trend analysis (Cheng Metro Line 9): Shihong Avenue - Shiling Forest Park Station - Shiling Station - fanjiayuanzi Station - east Beihu Park Station - shuangbayan Station - North Third Ring Station - Gu village - Shangyou station? Analysis of the original planning R6 trend of Luodai Town Station Tongan North Road Station Tongan station Cheng Chongqing Expressway intersection Station - shufangcun Park Station - Pucao Station - Longquan West Station - old Chengyu intersection Station - Economic Development Zone Station - Economic Development Zone South Station - Longquan Instrial Park Station - Shuanglu Road Station - Xinhua Road Station - Convention and Exhibition Center Station - Desai 2nd Street Station - Sanjiang Station - Desai 4th Street Station - New School East station of Sichuan University - Information College Station - Changcheng Road Station - reception Temple Station - yingchunqiao Station - Pengzhen North Station - Yanhe she station District Station (set R6 parking lot)
utility software in the Attached CD-ROM, and then connect it with USB cable to shoot successfully.
1
torch calls himself numpy in the field of neural networks, because he can put the tensor generated by torch in the GPU to speed up the operation (if you have a suitable GPU), just like numpy will put array in the CPU to speed up the operation. Free switch between torch and numpy:
Import torch import numpy as NP_ data = np.arange(6).reshape((2, 3)) torch_ data = torch.from_ numpy(np_ data) tensor2array = torch_ data.numpy() print( '&# 92; nnumpy array:', np_ data, # [[0 1 2], [3 4 5]] '&# 92; ntorch tensor:', torch_ data, # 0 1 2 \ n 3 4 5 [torch.LongTensor of size 2x3] '&# 92; ntensor to array:', Tensor2array, # [[0.12], [3.45]]) 1234567891011
mathematical calculation in Python: &# 160
many mathematical calculation functions in Python are the same as those in numpy
# ABS absolute value calculation data = [- 1, - 2,1,2] tensor = torch. Floattensor (data) # converted to 32-bit floating-point tensor print (&# 39&# 92; nabs', &# 39;&# 92; nnumpy: ', np.abs(data), # [1 2 1 2] '&# 92; ntorch: ', Torch. ABS (tensor) # [1 2] # sin trigonometric function
2. Variable Variables & # 160
the variable of Python is equivalent to a wrapper. If you want to transfer data to the graph built by python, you need to package the data with variable first. The packaged variable has three attributes: data, creator and grad
the document seems complete, what each layer does and what each parameter is written, but I don't read the code, In fact, many people can't understand its specific usage from the document. This can be seen from the discussion in issue. Similarly, there seem to be many examples, which can run directly and are all real world datasets. It seems very good, but in fact, for novices, if the model they need is not exactly the same as that in the example, it is not easy to understand what format the input and output data need to be. For example, examples do classification, but not sequence labeling. If you want to do POS tagging, you don't know how to organize the data. Of course, these problems can be solved by taking a day to read the code or go through the issue discussion. However, I believe many people will not read the code or read the discussion seriously, but change the tool directly. I feel that the current doc can only be understood by those who understand the code. It's useless if they don't understand the document
2. The project is very simple, so there are not many developers, but they are very active, and new things are added every day. Today, a new branch has been added. The backend can use theano or tensorflow. However, it seems that because scan is not supported, the backend does not implement recurrent layer with tensorflow. They are also aware of the document problem and feel that they need to add more tutorial for Xiao users instead of just showing it to developer
I haven't used any other frameworks, just keras & lt/ ol> It's very good to learn the basic usage of theano.
the code of the library itself is relatively simple and easy to read. As a novice in Python, I can also understand it. At present, there are two kinds of models: sequential and grapgh. The former does not mean recurrent, but means that the network is a stack of layers (including recurrent). Other main concepts include layer, regulator, optimizer and objective. Layer is used to build the output function of each layer. The model will use the output of the last layer to determine the final cost according to the objective and the regulator of each layer, and then use the optimizer to update the parameters ring update. If you look at these four and add the fit function in the model, you will use theano. Many models can cover, and seq2seq is also available. It is suggested that we not only look at examples, but also take a look at the issues discussion on GitHub. If we can't find it, we should ask questions directly. In terms of efficiency, I don't know how to optimize theano. I feel that this kind of package of keras has no cost, which is the same as using native theano. Of course, theano itself is so slow.. I don't know how to use it..