The Classification Problem

Describing insufficient data especially at the important point which are non-standard how you solve that problem which may be non-trivial so that others can actually replicate your result results ok we can kind of go through that or skip that a little bit and one thing which I said before which I wanted to mention is especially if you are kind of new to scientific writing start with this top-down approach give a global picture of what you want to communicate and once the global picture is given you can go deep in the individual for the individual components or aspects of your problem because you don’t do that if you start providing all the details from the beginning on the readers very very quickly lost and you may start with for example top level first explaining what the individual tasks are that you separate into component that you now have. Get to know more on classification onĀ Edusson.

Let’s say three components that you want to solve for example if you’re in the classification problem you have three problems the first problem is how to get your training data the second one may be how to get good features and the third one is how to realize a good classification system from technical perspective so you broke that down to three components and then you can describe the components with more details so explain the concept for let’s say the individual components like the classification system what is the underlying math behind that and then in the end at the lowest bottom provide the technical details it allows others to replicate your results okay experimental section again you want to describe what experiments have you conducted why the experiments have been conducted so the perfect case every block starts a sentence the first experiment is designed to show that the second experiment is designed to show that whatever this new sensor is 10 to 100 times more accurate than existing technology at the same time it is easier to use an existing technology then it means you would have to probe if you make these claims you will have to provide two experiments or at least to one showing the accuracy and the other one showing the ease of use.

If you don’t do that people will complain that you make claims that you which are not backed up by your experimental evaluation and that’s hard to argue against that so something which can actually kill a good technical paper if the claims that it makes are not supported by the experiments it always helps to compare your approach to other techniques so if you have a baseline technique and your new great method compared to what’s out there like in the report compare the nearest neighbor approach to the decision tree for example so make a comparison say which one is better what are the advantages of the individual messes yeah there again some example from the paper was indicating the divisional aspects what has been done and then we can have the conclusion is again a shorter part of the text.