diff options
| -rw-r--r-- | paper/figures/kronecker_l2_norm.pdf | bin | 0 -> 30708 bytes | |||
| -rw-r--r-- | paper/figures/kronecker_l2_norm_nonsparse.pdf | bin | 0 -> 30729 bytes | |||
| -rw-r--r-- | paper/figures/watts_strogatz.pdf | bin | 13567 -> 30785 bytes | |||
| -rw-r--r-- | paper/paper.tex | 2 | ||||
| -rw-r--r-- | paper/sections/experiments.tex | 19 | ||||
| -rw-r--r-- | src/convex_optimization.py | 2 | ||||
| -rw-r--r-- | src/make_plots.py | 60 |
7 files changed, 69 insertions, 14 deletions
diff --git a/paper/figures/kronecker_l2_norm.pdf b/paper/figures/kronecker_l2_norm.pdf Binary files differnew file mode 100644 index 0000000..5177233 --- /dev/null +++ b/paper/figures/kronecker_l2_norm.pdf diff --git a/paper/figures/kronecker_l2_norm_nonsparse.pdf b/paper/figures/kronecker_l2_norm_nonsparse.pdf Binary files differnew file mode 100644 index 0000000..18ceabf --- /dev/null +++ b/paper/figures/kronecker_l2_norm_nonsparse.pdf diff --git a/paper/figures/watts_strogatz.pdf b/paper/figures/watts_strogatz.pdf Binary files differindex 79df8d4..ddaa525 100644 --- a/paper/figures/watts_strogatz.pdf +++ b/paper/figures/watts_strogatz.pdf diff --git a/paper/paper.tex b/paper/paper.tex index 1201b0f..36e87be 100644 --- a/paper/paper.tex +++ b/paper/paper.tex @@ -41,7 +41,7 @@ % note in the first column to ``Proceedings of the...'' %\usepackage[accepted]{icml2015} \usepackage[utf8]{inputenc} - +\usepackage{caption} % The \icmltitle you define below is probably too long as a header. % Therefore, a short form for the running title is supplied here: \icmltitlerunning{Sparse Recovery for Graph Inference} diff --git a/paper/sections/experiments.tex b/paper/sections/experiments.tex index 600cac3..ccd82ce 100644 --- a/paper/sections/experiments.tex +++ b/paper/sections/experiments.tex @@ -3,15 +3,18 @@ \caption{Precision-Recall curve Holme-Kim Model. 200 nodes, 16200 edges.} \end{figure} -\begin{figure} -\includegraphics[scale=.4]{figures/watts_strogatz.pdf} -\caption{Watts-Strogatz Model. 200 nodes, 20000 edges.} -\end{figure} +\begin{table*}[t] +\centering +\begin{tabular}{c c c c} -\begin{figure} -\includegraphics[scale=.4]{figures/barabasi_albert.pdf} -\caption{Barabasi Model.} -\end{figure} +\includegraphics[scale=.21]{figures/barabasi_albert.pdf} +& \includegraphics[scale=.21]{figures/watts_strogatz.pdf} +& \includegraphics[scale=.23]{figures/kronecker_l2_norm.pdf} +& \includegraphics[scale=.23]{figures/kronecker_l2_norm_nonsparse.pdf}\\ +(a) & (b) & (c) & (d) +\end{tabular} +\captionof{figure}{blabla} +\end{table*} In this section, we validate empirically the results and assumptions of Section~\ref{sec:results} for different initializations of parameters ($n$, $m$, $\lambda$) and for varying levels of sparsity. We compare our algorithm to two different state-of-the-art algorithms: \textsc{greedy} and \textsc{mle} from \cite{Netrapalli:2012}. As an extra benchmark, we also introduce a new algorithm \textsc{lasso}, which approximates our \textsc{sparse mle} algorithm. We find empirically that \textsc{lasso} is highly robust, and can be computed more efficiently than both \textsc{mle} and \textsc{sparse mle} without sacrificing for performance. diff --git a/src/convex_optimization.py b/src/convex_optimization.py index 8dc6f82..0d506e1 100644 --- a/src/convex_optimization.py +++ b/src/convex_optimization.py @@ -90,7 +90,7 @@ def diff_and_opt(M_val, w_val, f_x, f_xz): def F(x=None, z=None): if x is None: - return 0, cvxopt.matrix(-.001, (n,1)) + return 0, cvxopt.matrix(-.1, (n,1)) elif z is None: y, y_diff = f_x(x, M_val, w_val) return cvxopt.matrix(float(y), (1, 1)),\ diff --git a/src/make_plots.py b/src/make_plots.py index d83eb2c..792e325 100644 --- a/src/make_plots.py +++ b/src/make_plots.py @@ -108,6 +108,58 @@ def plot_barabasi_albert_graph(): +def plot_kronecker_l2norm(): + plt.clf() + fig = plt.figure(1) + x = [50, 100, 500, 1000, 2000] + sparse_recov = [62, 60, 36, 28, 21] + max_likel = [139, 101, 42, 31, 25] + lasso = [50, 48, 33, 29, 23] + + fig, ax = plt.subplots() + + plt.subplots_adjust(bottom=.2, top=.85) + plt.xticks(ha="right", rotation=45) + + plt.axis((50, 2000, 0, 145)) + plt.xlabel("Number of Cascades") + plt.ylabel("l2-norm") + plt.grid(color="lightgrey") + ax.plot(x, lasso, 'ko-', color="orange", label="Lasso") + ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE") + ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method") + plt.legend(loc="upper right") + ax.set_xticks(x) + ax.set_xticklabels(tuple(x)) + plt.savefig("../paper/figures/"+"kronecker_l2_norm.pdf") + + +def plot_kronecker_l2norm_nonsparse(): + plt.clf() + fig = plt.figure(1) + x = [50, 100, 500, 1000, 2000] + sparse_recov = [56, 55, 28, 21, 15] + max_likel = [125, 80, 35, 25, 20] + lasso = [47, 47, 27, 22, 17] + + fig, ax = plt.subplots() + + plt.subplots_adjust(bottom=.2, top=.85) + plt.xticks(ha="right", rotation=45) + + plt.axis((50, 2000, 0, 145)) + plt.xlabel("Number of Cascades") + plt.ylabel("l2-norm") + plt.grid(color="lightgrey") + ax.plot(x, lasso, 'ko-', color="orange", label="Lasso") + ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE") + ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method") + plt.legend(loc="upper right") + ax.set_xticks(x) + ax.set_xticklabels(tuple(x)) + plt.savefig("../paper/figures/"+"kronecker_l2_norm_nonsparse.pdf") + + def plot_ROC_curve(figure_name): """ plot information in a pretty way @@ -168,8 +220,8 @@ if __name__=="__main__": #convex_optimization.type_lasso) if 1: compute_graph("../datasets/kronecker_graph_256_cross.txt", - n_cascades=2000, lbda=0.1, min_proba=.2, max_proba=.7, + n_cascades=50, lbda=0., min_proba=.2, max_proba=.7, passed_function= - #convex_optimization.sparse_recovery, - convex_optimization.type_lasso, - sparse_edges=False)
\ No newline at end of file + convex_optimization.sparse_recovery, + #convex_optimization.type_lasso, + sparse_edges=True)
\ No newline at end of file |