Tips for choosing a paper for the report :
Strategy 1: `do what you love'
One strategy is to choose a subject you're interested in, even if not addressed in class, and even if you're not sure some modeling work exists in this area. For example, one student told me he would like to do something on migraine. Other examples could be: dyslexia, sleep, face recognition, mathematical ability, alzheimer's disease, depression, addiction, fear... whatever intrigues you.
The cognitive aspect means that it has to be related to perception, attention, memory, learning, language, decision making, mental illness, emotions, consciousness... Not so well defined... Please, ask me if you're not sure.
Of course topics addressed in the lectures are fine too (and will be somewhat easier).
Then, investigate what models exist in this area. For this, use the web, pubmed, and/or ask someone who might know, e.g. someone who knows well the experimental side of that domain.
You'll often find that not so many models have been proposed. And even fewer good models. Indeed, the field is in its very infancy (which is why we need you !).
A lot of work in computational neuroscience is not cognitive because it deals with describing circuits and dynamics without explicity putting them in relation with mental phenomena (e.g. a model describing the formation of maps in visual cortex, or a new fancy model of a neuron, or a model showing how synaptic delays can influence the dynamics of the circuit).
What's 'computational' is also not very well defined. It can be more 'theoretical' than computational (some maths, but no simulations). It doesn't need to be a 'neural network model'. Mathematical models, statistical models, machine learning techniques, information theory analysis can be fine, as long as they are not trivial, and can complement experiments.
It can even be an experimental paper, as long as the experiments are guided by theory, or interpreted in a theoretical framework.
A good model is one that helps to describe the data, test hypothesis, make predictions.
If you don't find anything interesting in the area of your choice, you can then settle for a related question that has been a subject of theoretical work, for e.g. the student interested in migraine will find that one particular aspect of migraines, the visual hallucinations occurring during the migraine aura, have been studied and modeled [see eg. Reggia 1996].
If you feel very inspired, you can also discuss how some modeling tools that have been developped for another problem could be applied to the area of your choice.
In any case, you can enrich your report by discussing i) what questions models can address (in this area), ii) the limitations of the currents models and reasons for limitations, iii) ways to improve current models, or apply known techniques to new area.
I will take into account the difficulty of the chosen article(s)/theme(s).
It's okay if the paper is not recent (but then make sure to cite some recent papers in your discussion).
I encourage papers that deal explicitly with the relationship between the physiological substrate (neural activity) and some behavioral/ psychophysical/cognitive aspect.
Curiosity is encouraged.
Strategy 2: `follow the stars'
An easier route is to check what the leading crowd does.
A combination of these two strategies is probably a good idea.