Record their growth, in order to urge their own, but also to help others.
I will live a focused life, because this is the best choice.
Benchmarking research study: I will get national award, received his doctorate in American universities offer; the depth of daily working hours in 12h
is not the schedule, the work is complete and record the depth of working time
strive to become good people, become credibility high man, only to become such a person, in order to maximize their influence in the form of writing, lectures and other
title: the number of weeks, the depth of working hours, when the main task of the week
the depth of daily working hours (research study)
Tuesday, 6.5h
morning: configure the required test software environment, download the appropriate installer and python libraries: Keras, Tensorflow, Cuda, cuDNN , 2.5h
afternoon: Use Keras implements a simple convolution neural network, 1h; Mathematical Statistics course talked parameter estimation and hypothesis testing in multivariate normal distribution, as well as some questions over the years, 0.5h
at night: on the server and personal PC laboratory to reproduce the experimental results are handwriting recognition, 0.5h
Wednesday, 3.5h
morning: develop a reproducible experimental program, 0.5h
afternoon: a paper reproduction of experiments: experiments ST-ResNet papers in New York bicycle dataset, 1h
night: watched two papers: ST- ResNet, 2h
Thursday, 6h
am: reading the paper frame content model can explain the sketch: ST-ResNet paper, 1.5h
pm: Start determines the input format model: ST-ResNet based on meshing bicycle data sets, 2h
night: determining the input format and the model begins processing the original data set: subway stations sequentially input mesh, 2.5h
Friday, 9.5h
am: start processing the raw data set; practice questions over the years on Mathematical Statistics Course, 3.5h
afternoon: continue processing the original data set; 3h
at night: h5 generate the required model input file; and modify the source code, start training model; 3h
Saturday, 6.5h
morning: 2 to determine the operating range of topics paper, 2h
pm: Reference Papers: DeepPF, time frame modeling and model space, subway operating information; determining the main content of the thesis: DNN, CNN, RNN; 2h
pm: super start writing parameters and validation sets: reference "deep learning", that took the book; and by modifying the parameters of super-group made 9 ST-ResNet experiment; 2.5h
Sunday, 6h
am: start doing their homework introduction part 2, 2h
pm: Introduction to finish the job part 2, 1h
night: 1-7 blooms, deep understanding of the development process associated with deep learning of basic concepts, 3h
Weekly summary
Far from reaching the target depth work 12 hours a day, and in one sentence to motivate themselves (learned from around ZC classmates):
Do not lose confidence and hope, do not escape, start from whatever little things, do not hurry, take your time, it's just a short way of life in the marathon, it just a little difficult.