Try HEFALMP (http://sonorus.princeton.edu/hefalmp/). It's a naive bayesian classifier based on co-expression and physical interaction data. It creates highly informative networks.

EnrichR is a suite that calculates enrichments of any sort and also creates networks using several tools like Reactome and others (http://amp.pharm.mssm.edu/Enrichr/)

@Ulrich and @Sachin Can these tools be used for plant systems as well?

Can I ask, do you want to infer a network from expression data, or do you want to take an existing network and subsample it?

Network Work Bench (freely downloadable on the Web) is the Barabasi group software and works very well.

I have actually calculated the expression correlations. I just want to provide a suitable visualization tool which can use the expression correlation values as weights

You could try GraphViz. It takes a text file describing a network and a bunch of other command line parameters and plots a graph. I haven't tried it, but I've heard of it being used on websites to plot graphs. Alternatively, you could use something in javascript like the CytoscapeWeb framework, or D3.

HEFALMP is for homo sapiens only, I'm not sure for EnrichR. In alternative you can use DAVID (http://david.abcc.ncifcrf.gov/tools.jsp) which can be used for Arabidopsis at least.

But if you have expression correlations (between how many samples?) and you just which to visualize these correlations a simple heatmap might do it, especially if you wish to cluster samples with similar expression correlations. Just use the correlation values instead of the expression values and perform hierarchical clustering (using for example TMEV http://www.tm4.org/).