Attend meet-ups, and be more social with other developers

Embrace the reactive programming approach

I can’t stress enough how important this is. You can’t consider a feature to be complete until you have written exhaustive test cases for it. Tests will help you build confidence walls across your code.

Having a playground project in progress can completely change the way you learn things. If you come across something new in Android somewhere, make it a habit to immediately give it a try in your playground project.

This will change the way you look at the apps you have been coding. Try to talk and interact more with the UI and UX designers in your team to get better insight into app design.

Make it a habit to discover new and better ways of making your tools work for you, thereby improving your workflow and productivity.

If you have developed a library, plugin, or other useful piece of code and you're using it in your own app, consider open-sourcing it. There's much to learn in the process of contributing to open-source projects or maintaining your own. It's an excellent crash course in open-source development that will exponentially increase your value as a developer.

I can’t stress enough how important this can be in your Android development career. Whenever you are stuck in trying to solve a critical programming problem, design patterns can be a lifesaver.

But that will not help you grow as a complete Android developer. The only way to truly become a better developer is to read the excellent code of more experienced developers. You should start looking at other open-source apps and libraries, where you'll discover a lot of coding techniques and feature implementations that you had no idea about before.

need

It does not matter what is the platform you code on; GIT version control is something you will use constantly in every decent company or project of your own. You can never be too good at GIT. You probably can never master it as well. This thing is huge. Just try to use it as efficiently as possible. It’s especially important when working in team.

If you decided to get into some development team, try to find one with at least one developer you consider better than yourself. He is the one you are going to learn most from. No matter how experienced you are, this job is about constant development of yourself as well. It’s good to have some mentor of your own.

As a mobile developmer you might often work alone. That’s because many projects aren’t big enough to throw more people into it. Some of you will actually like it but what I suggest is at least to try getting into some development team. It’s way more fun this way. And more important: You learn much more from the project.

The coding is about writing something and then spending most of the time making it work.

Command + Option + O

针对接口编程，不针对实现

多用组合，少用继承（我们使用组合的方式，为客户设置了算法）

策略模式（Strategy Pattern）：定义了算法族，分别封装起来，让它们之间可相互替换，此模式让算法的变化独立于使用算法的客户。

定义了对象之间的一对多的依赖，这样一来，当一个对象改变时，它的所有的依赖者都会收到通知并自动更新。

而且也没有办法使用组合的模式实现这三个方法的复用

对于JDK或者Andorid中都有很多地方实现了观察者模式，比如XXXView.addXXXListenter ， 当然了 XXXView.setOnXXXListener不一定是观察者模式，因为观察者模式是一种一对多的关系，对于setXXXListener是1对1的关系，应该叫回调。

在医学图像分析诊断预测里面，非常重要一点是模型的可解释性。就是你得解释清楚你这个AI模型为什么work了，或者为什么在某些情形不work。

由于没有像CV里面物体识别场景识别里标准化的数据库和评测方法，一直比较小众。

学位论文的开题报告需要有选题、问题、文献综述、内容、方法、理论框架等内容， 而所有这些内容具有内在的逻辑一致性， 选题要具有意义必须通过问题的研究和解决来体现，而文献综述是围绕着提出的问题来展开的，研究的内容是根据文献综述和研究问题来确定的，研究方法一定是从属于研究的问题和研究内容的，而理论框架、分析视角是为了使方法使用的有效性得以展现。

Cookies are small piece of information that is sent by web server in response header and gets stored in the browser cookies. When client make further request, it adds the cookie to the request header and we can utilize it to keep track of the session. We can maintain a session with cookies but if the client disables the cookies, then it won’t work.

is a conversional state between client and server and it can consists of multiple request and response between client and server. Since HTTP and Web Server both are stateless, the only way to maintain a session is when some unique information about the session (session id) is passed between server and client in every request and response.

Artboards behave like a special kind of group. The main differences are that they are represented on the Canvas, you never have to double-click to view their contents, and they don’t resize to fit the contents inside.

Layers in your document can be grouped so they will appear as one layer.

Take the example of a flower; create just one leaf shape, then rotate a dozen copies around the center and you’ll have your flower.

n the first phase, an XML document is partitioned into chunks and parsedin parallel. In the second phase, partial DOM node tree structures created duringthe first phase, are linked together (in parallel) to build a complete DOM nodetree.P

singular or plural

such as using previous video frames might inform the understanding of the present frame. If RNNs could do this, they’d be extremely useful. But can they? It depends.

A recurrent neural network can be thought of as multiple copies of the same network, each passing a message to a successor. Consider what happens if we unroll the loop

如果是繁重工作需要大量消耗电量，应该插电源。 

建议多用电池，不建议一直插着电源，不是说一直插着电源就会烧坏电脑,

iPhone如果充到100%就拔掉，和充满后再过半个小时拔掉，明显后者用的时间更久。

而且建议电量低于20%就要充,因为电池电量低,电压也会低,长期的低电压状态会对锂电池造成危害.

把自己的代码全部放入唯一的名字空间MYAPP中，会大大减少全局变量冲突的可能。

虽然是strict模式，但语句var x = 'Hello, ' + y;并不报错，原因是变量y在稍后申明了。但是alert显示Hello, undefined，说明变量y的值为undefined。这正是因为JavaScript引擎自动提升了变量y的声明，但不会提升变量y的赋值。

有点像您可以在不同 Activity 中重复使用的“子 Activity”

Whenever there were more than two cavities, it wasn’t clear anymore if that cavity was part of the lung.

In our approach blobs are detected using the Difference of Gaussian (DoG) method, which uses a less computational intensive approximation of the Laplacian operator. We used the implementation available in skimage package.

rotation augmentation.

It behaves well for the imbalance that occurs when training on smaller nodules, which are important for early stage cancer detection.

this causes a difference in spacing between voxels of the original scan.

We rescaled and interpolated all CT scans so that each voxel represents a 1x1x1 mm cube.

We used this dataset extensively in our approach, because it contains detailed annotations from radiologists.

有超强的团队精神、专业的知识背景、良好的沟通技巧，这些用烂的词语，HR每天要看几百遍，还会有一丢丢感觉吗？

正确的写法是“姓名+应聘岗位+工作年限”，例如“宋小宝-新媒体运营-3年公众号运营经验”。 如果是应届毕业生，可以写成“院校+姓名+应聘岗位”。

For either case, the second argument is what we want to see as output when we apply the function. f may then be used like a normal Python function.

symbols (Variables)

It is good to think of theano.function as the interface to a compiler which builds a callable object from a purely symbolic graph. One of Theano’s most important features is that theano.function can optimize a graph and even compile some or all of it into native machine instructions.

A perceptron takes several binary inputs, x1,x2,…x1,x2,…x_1, x_2, \ldots, and produces a single binary output

检测任务这个方法显然并不适合，因为原始图像如果缩小到224这种分辨率，那么感兴趣对象可能都会变的太小无法辨认。

region进行特征提取时会有重复计算，Fast-RCNN正是为了解决这个问题诞生的

在CNN火起来之前，对象检测这一问题基本是遵循着“设计手工特征(Hand-crafted feature)+分类器”的思路

RCNN 解决的是，“为什么不用CNN做classification呢？” Fast-RCNN 解决的是，“为什么不一起输出bounding box和label呢？” Faster-RCNN 解决的是，“为什么还要用selective search呢？”

end-to-end

The masks are constructed by using the diameters in the nodule annotations.

Since the nodule segmentation network could not see a global context, it produced many false positives outside the lungs, which were picked up in the later stages. To alleviate this problem, we used a hand-engineered lung segmentation method.