An important part of research is awareness of the context of your project, and in particular of related work done by others. This involves finding out about the existence of relevant publications; and, once you've unearthed their details, actually getting hold of them.
You can also find a much larger range of items using the Everything search. This will pull up individual journal and conference proceedings articles, but also preprints on arXiv or e.g. papers that quote authors you have searched for. It is therefore a good way of ensuring you don't miss anything, though the variety of item formats can be a little awkward to organize.
For something slightly more focussed, with support for marking items, systematic forward and backward (see below) reference searches, BibTeX output etc. I would recommend Web of Knowledge, which by default goes through to Web of Science as its main database. This and other databases require a login, which is normally the same as your KCL email username and password (see here).
Here's a brief introduction to Web of Knowledge: clicking on ``click here to access Web of Science'' gets you to the login screen; once you're past this, leave the default option ``Web of Science'' or click on the ``All databases'' tab for other options. Web of Science is a very comprehensive database that indexes journal publications from about 1981 onwards, though historical coverage is expanding steadily. If e.g. you want to do an author search, enter ``coolen acc'' in the relevant field and click ``search'' and you will get a list of all of Prof Coolen's journal publications. Note that the initials matter: e.g. ``coolen ac'' will find only a few articles; ``coolen a'' will find also articles by a different person; ``coolen'' locates articles by anyone with that last name. You can also search for author combinations, e.g. ``coolen and sollich'' (or ``coolen acc and sollich p'') will show you that Prof Coolen and I have written some papers together.
By searching for a topic rather than an author you can find articles on specific topics; e.g. the search term ``complex and non-equilibrium'' including the quotes will show you all articles containing the three words as a phrase in their title, abstract or keywords. If you omit the quotes, all articles come up that contain the three words in any order and whether together or not - too many!
You can use asterisks ``
'' as wildcards in the usual way, i.e.
``comp'' picks up complex, computational, comparison etc.
You can also search by publication (e.g. journal) name, date, or
title (in which case abstracts are not searched).
Once you have found a few articles of interest, a good way of finding other work in the same area is to look for citations: from any given article, you can get a list of all papers which are referred to by clicking on ``cited references'' on the right hand side. In this way, you can go backwards through the literature. On the other hand, by clicking on ``times cited'' you will get all papers referring to the article you're looking at. This lets you look forward, towards recent work. It's particularly useful once you've hit on a seminal paper in a given area, which most of the later work on the subject is likely to cite.
During all this searching, you'll want to keep the most relevant articles. This is done by ``marking'': if you're on a screen showing details of a single article, there should be a ``add to marked list'' button; for lists of articles you can mark them by ticking the boxes by the papers you are after and clicking ``submit marks'' once you've made your selection (don't forget this second step, otherwise the clicked boxes have no effect). Once you've marked some articles, the ``marked list'' item near the top of the screen will show how many articles you've marked. The most useful thing that you can do with the marked list is ``save to file'', which downloads the information about the articles in whatever format you've specified (including various items such as the abstracts; these items are specified by ticking the relevant boxes on the marked list screen). You can cut-and-paste this info from the downloaded file directly into your reference list; see later. Saving to Endnote, RefMan etc is also available.
By default, searches from the Web of Knowledge home page access the Web of Science database. Other options include ISI proceedings. The latter provides similar information to Web of Science, but for articles published in conference proceedings. The coverage here is, however, somewhat patchier than for Web of Science, and there are some proceedings volumes that are not catalogued here. Often, e.g. if you are mainly interested in journal articles anyway, it is useful to use only the Web of Science database by clicking on the appropriate tab. Otherwise an ``all database'' search can show you several occurrences of the same article as separate items, because the information is drawn from different databases. This can also make dealing with marked lists more complicated as these are then split according to what database the information comes from.
Apart from searching such general literature databases, there are also some sites which are specific to certain research areas. Citeseer is a big database of articles in computer science, where many papers on the machine learning and statistical inference aspects of complex systems modelling; as on Web of Science, you can search for papers either by certain authors or quoting work of certain authors. For each paper there is normally an abstract and some reference information; much of the data is machine-generated, and so is not as accurate as the ``human-corrected'' data on Web of Science. Nevertheless, citeseer is a good resource; unlike Web of Science, it also has links to the texts of the actual papers in most cases (look at the small box in the top right hand corner, which offers various formats including PDF and postscript).
Not dissimilar is the Los Alamos preprint archive arXiv, which has a vast collection of papers in Mathematics, Physics and Computer Science; some complex systems-related papers are in the Disordered Systems and Neural Networks section of the condensed matter physics archive, others in the computer science archives. As on citeseer and elsewhere, you should be aware that the papers you get here are often ``preprints'', i.e. pre-publication versions. If/when the papers eventually get published in a peer-reviewed journal, the final version can be slightly different, but radical changes which are not recorded in the archive are rare.
There are also other, more specialized sites; e.g. the Kernel Machines web site is a good place to look for work on Gaussian processes and Support Vector Machines. Your supervisor may well be able to suggest others.
Finally, you can of course try relevant search terms on general internet search engines such as google or more specifically google scholar. This tends to be more hit-and-miss and the results will need more sifting. It can nevertheless be useful, e.g. for finding people who have worked in the area you're interested in; you can then get more information from their home page or look for their work in literature databases. Internet search engines are also a good way of looking for books - which are not normally indexed in journal and proceedings publication databases - that are not in the KCL library catalogue.
Let's say you've now located the references to the ``hot'' articles in the area of your project; how do you actually get hold of them? You may be lucky and have already found a full-text version of the papers when you were looking on the internet. Otherwise: