As you loop over the folds you produce performance values,  for example Accuracy, Precision and Recall. These values vary a little bit depending on how you slice the data for each fold.

The range of these values that you obtain represents the range of variability in the performance of your model that you might expect to see if you brought that model to be tested upon a brand new data set.

From the 10 performance metrics that you obtained for each classifier you can calculate a mean and a corrected sample standard deviation.  Now you would like to know for two different classifiers with two different average performance values, whether those differences are “significant” (i.e. whether the differences are meaningfully different, or are simply due to the same random statistical fluctuations that account for the same fold to fold random variation that has been observed in the first place.

If you have a performance metric average A1 and a standard deviation std1 for a classifier C1 and the same for another classifier C2. You can estimate whether the difference between their relative performance is meaningfully  different from 0 by calculating Student t test.

If the score is large ( usually a value greater than 3 standard deviations ) one may declare that the performance of the classifiers to be significantly different. If not, they are essentially equivalent.


Research Paper Parts

Posted: December 22, 2017 in Uncategorized

DISCLAIMER :::  This is my personal copy of notes extracted from . Visited in Dec/22/2017.

One of the most important aspects of science is ensuring that you get all the parts of the written research paper in the correct order.

You may have finished the best research project on earth but if you do not write an interesting and well laid out paper, then nobody is going to take your findings seriously.

The main thing to remember with any research paper is that it is based on an hourglass structure. It begins with general information and undertaking a literature review, and becomes a more specific as you nail down a research problem and hypothesis.

Finally, it again becomes more general as you try to apply your findings to the world at general.

Whilst there are a few differences between the various disciplines, with some fields placing more emphasis on certain parts than others, there is a basic underlying structure.

These steps are building bllocks of constructing a good research paper. This section outline how to lay out the parts of a research paper, including the various experimental methods and designs.

The principles for literature review and essays of all type follow the same basic principles.

  • Abstract
  • Introduction
  • Method
  • Results
  • Discussion
  • Conclusion
  • References


For many, writing the introduction is the first part of the process, setting down the direction of the paper and laying out exactly what the researcch paper is trying to achieve.

For others, the intriduction is the last thing written, acting as a quick summary of the paper. As long as you have planned a good structurefor the parts of a research paper, both approaches are acceptable and it is a master piece.

A good introduction generally consists of three distinct parts:

1. You should first give a general presentation of the research problem.

2. You should then lay out exactly what you are trying to achieve with this particular research project.

3. You should then state your own position.

Ideally, you should try to give each section its own paragraph, but this will vary given the overall length of the paper.


This should be the easiest part of the paper to write, as it is a run-down of the exact design and methodology used to perform the research. Obviously, the exact methodology varies depending upon the exact field and type of experiment.

There is a big methodological difference between the apparatus based research of the physical sciences and the methods and observaiton methods of social sciences. However, the key is to ensure that another researcher would be able to replicate the experiment to match yours as closely as possible, but still keeping the section concise.

You can assume that anybody reading your paper is familiar with the basic methods so try not to explain every last detail. For example, an organic chemist or biochemist will be familiar with chromatography, so you only need to highlight the type of equipment used rather than explaining the whole proces sin detail.

In the case of a survey, if you have too many questions to cover in the method, you can always include a copy of the questionnaire in the appendix. In this case, make sure that you refer to it.


This is probably the most variable part of any research paper, and depends on the results and aims of the experiment. For quantitative researcch, it is a presentation of the numerical results and data, whereas for qualitative researcch it should be a broader discussion of trends, without going into too much details.

For research generating a lot of results, then it is better to include tables or graphs of the analyzed data and leave the raw data in the appendix, so that a researccher can follow up and check your calculations.

A commentary is essential to linking the results together, rather than just displaying isolated and unconnected charts and figures.

It can be quite difficult to find a good balance between the results and the discussion section, because some findings, especially in a quantitative or descriptive experiment, will fall into a grey area. Try to avoid repeating yourself too often.

It is best to try to find a middle path, where you can give a general overview of the data and then expand on it in the discussion. You should try to keep your own opinions and interpretations out of the results section, saving that for the discussion later on.


This is where you elaborate on your findings and explain what you found, adding your own personal interpretations. Ideally you should link the discussion back to the introduction addressing each point individually. It is important to make sure that every piece of information in your discussion is directly related to the thesis statements or you risk cluttering your findings. In keeping with the hourglass principles, you can expand on the topic later in the conclusion.


The conclusion is where you build on your discussion and try to relate your finding to other research and to the world at large. In a short research paper it may be a paragraph or two or even a few lines.

In a dissertation it may well be the most important part of the entire paper – not only does it decribe the results and discussion in detail, it emphasizes the importance of the results in the field, and ties it in with the previous research.

Some research papers require a recommendations section, postulating the further directions of the research, as well as highlighting how any flaws affected the results. In this case you should suggest any improvements that could be made to the research design.

Research Hypothesis

Posted: December 22, 2017 in Uncategorized

Very nice instructions on how to write a research hypothesis


Collected from  on 22/12/2017


Posted: August 15, 2017 in Uncategorized
$ ssh-add -l
$ ssh-add ~/.ssh/id_rsa

After trying to upgrade Fedora 24 to Fedora 25  using  dnf system-upgrade the system went into a rebooting loop showing the following errors :

mount: /oldsys/sys: filesystem mounted, but mount(8) failed: Permission denied
mount: /oldsys/proc: filesystem mounted, but mount(8) failed: Permission denied
mount: /oldsys/run: filesystem mounted, but mount(8) failed: Permission denied
mount: /oldsys/dev: filesystem mounted, but mount(8) failed: Permission denied
dracut Warning: Killing all removing processes
Kernel not configured for semaphores ( System V IPC). Not using udev synchronisation code.
reboot: Restarting system

To solve the issue I edited GRUB menu during boot time and added the following two parameters to the end of the string starting with “linux /boot/vmlinuz-linux”  selinux=1 enforcing=0 

As simple as that :

$ A  \cup B $   ==  A \cup B

$ A  \cap  B $  ==  A \cap B

Site com novo visual