Large Language Models (LLMs) are inherently constrained by their fixed-length context windows, which limits LLMs' ability to retain and utilize information across long-term interactions.To address this limitation, recent work has proposed external memory modules for LLMs.Using memory modules typically involves two stages: evidence retrieval and memory utilization.While prior work focuses on the architecture of memory modules and the retrieval stage, the equally critical memory utilization stage remains underexplored.Building on this, we propose MemCoRL, a two-stage alternating co-optimization reinforcement learning method.Stage 1 optimizes evidence retrieval using citation feedback and semantic accuracy from utilization as rewards.Stage 2 optimizes utilization with rewards combining semantic similarity and lexical overlap.Iterative co-optimization establishes a positive feedback loop: better retrieval improves memory utilization, which in turn refines retrieval rewards.Experimental results show our approach outperforms the leading baselines on both lexical overlap and semantic similarity metrics, confirming the co-optimization in memory retrieval and memory utilization.

In high-dimensional data clustering practices, the cluster structure is commonly assumed to be confined to a limited number of relevant features, rather than the entire feature set. However, for high-dimensional data, identifying the relevant features and discovering the cluster structure are still challenging problems. To solve these problems, this paper proposes a novel fuzzy c-means (FCM) model with sparse regularization (ℓq(0 <; q ≤ 1)-norm regularization), by reformulating the FCM objective function into the weighted between-cluster sum of square form and imposing the sparse regularization on the weights. An algorithm is also developed to explicitly solve the proposed model. Compared with the existing clustering models, the proposed model can shrink the weights of irrelevant features (noisy features) to exact zero, and also can be efficiently solved in analytic forms when q = 1,1/2. Experiments on both synthetic and real-world data sets show that the proposed approach outperforms the existing clustering approaches.