fixing test

src/algorithms/stochastic_rk_means.jl

@@ -14,14 +14,9 @@ run_algorithm(::Type{DROP_OUT}, X, Y, dfunc::Function, alg_params::Vector, k::In
 run_algorithm(::Type{ADA_L1RDA}, X, Y, dfunc::Function, alg_params::Vector, k::Int, max_iter::Int, tolerance::Float64, online_pass, train_idx) = 
 			  adaptive_l1rda_alg(dfunc, X, Y, alg_params..., k, max_iter, tolerance, online_pass, train_idx)
 
-function At_mul_B!(C::Array{Float64,2}, A::SparseMatrixCSC, B::SparseMatrixCSC)
-    assert(size(A,1) == size(B,1))
-    C[:,:] = At_mul_B(A,B)
-end
-
-function At_mul_B!(C::Array{Float64,2}, A::SparseMatrixCSC, B::Array{Float64,2})
-    At_mul_B!(C,full(A),B)
-end
+At_mul_B!(C::Array{Float64,2}, A::SparseMatrixCSC, B::SparseMatrixCSC) = begin C[:,:] = At_mul_B(A,B) end
+At_mul_B!(C::Array{Float64,2}, A::SparseMatrixCSC, B::Array{Float64,2}) = At_mul_B!(C,full(A),B)
+At_mul_B!(C::Array{Float64,2}, A::Array{Float64,2}, B::SparseMatrixCSC) = At_mul_B!(C,A,full(B))
 
 # core algorithmic part
 function stochastic_rk_means{A <: Algorithm}(X, rk_means::RK_MEANS{A}, alg_params::Vector, k::Int, max_iter::Int, 
@@ -64,14 +59,15 @@ function stochastic_rk_means{A <: Algorithm}(X, rk_means::RK_MEANS{A}, alg_param
 
     	result = @parallel (hcat) for cluster_id in unique(mappings)
     		cluster_idx = find(mappings .== cluster_id)
-    		run_algorithm(rk_means.support_alg,X,Y,dfunc,alg_params,k,max_iter,tolerance,online_pass,cluster_idx)[1]
+    		r = run_algorithm(rk_means.support_alg,X,Y,dfunc,alg_params,k,max_iter,tolerance,online_pass,cluster_idx)[1]
+            size(r) == (d,) ? r'' : r
     	end
 
     	# assign and check the result of parallel execution
     	if all(result .== 0) || all(isnan(result))
     		failed_mapping = true
     		w = rand(d,rk_means.k_clusters)
-    	elseif size(result,2) != size(w,2)
+    	elseif size(result) != size(w)
     		failed_mapping = true
     		diff = size(w,2) - size(result,2) 
     		w = [result rand(d,diff)]