diff --git a/btas/generic/cp_df_als.h b/btas/generic/cp_df_als.h
index dd8b8c68..d26c223f 100644
--- a/btas/generic/cp_df_als.h
+++ b/btas/generic/cp_df_als.h
@@ -672,9 +672,9 @@ namespace btas {
     /// return if \c fast_pI was successful
     void build_random(ind_t rank, ConvClass &converge_test, bool direct, ind_t max_als, bool calculate_epsilon,
                       double &epsilon, bool &fast_pI) override {
+      boost::random::mt19937 generator(random_seed_accessor());
+      boost::random::uniform_real_distribution<> distribution(-1.0, 1.0);
       if(!factors_set) {
-        boost::random::mt19937 generator(random_seed_accessor());
-        boost::random::uniform_real_distribution<> distribution(-1.0, 1.0);
         for (size_t i = 1; i < ndimL; ++i) {
           auto &tensor_ref = tensor_ref_left;
           Tensor a(Range{Range1{tensor_ref.extent(i)}, Range1{rank}});
@@ -693,6 +693,36 @@ namespace btas {
           this->A.push_back(a);
         }
 
+        Tensor lambda(rank);
+        lambda.fill(0.0);
+        this->A.push_back(lambda);
+        for (size_t i = 0; i < ndim; ++i) {
+          normCol(i);
+        }
+      } else{
+        for(size_t i = 0; i < this->ndim; ++i) {
+          auto &a_prev = this->A[i];
+          ind_t col_dim = a_prev.extent(0);
+          ind_t prev_rank = a_prev.extent(1), smaller_rank = (prev_rank < rank ? prev_rank : rank),
+                larger_rank = (smaller_rank == prev_rank ? rank : prev_rank);
+          if (larger_rank > smaller_rank) {
+            Tensor a(col_dim, larger_rank);
+            for (auto iter = a.begin(); iter != a.end(); ++iter) {
+              *(iter) = distribution(generator);
+            }
+            auto lo_bound = {0l, 0l}, up_bound = {col_dim, smaller_rank};
+            auto view = make_view(a.range().slice(lo_bound, up_bound), a.storage());
+            //std::copy(view.begin(), view.end(), a_prev.begin());
+            auto old = a_prev.begin();
+            for(auto iter = view.begin(); iter != view.end(); ++iter, ++old)
+              *(iter) += *(old);
+            a_prev = a;
+          } else{
+            for(auto & el : a_prev)
+              el +=  distribution(generator);
+          }
+        }
+        A.pop_back();
         Tensor lambda(rank);
         lambda.fill(0.0);
         this->A.push_back(lambda);