Learning What to Remember: Long-term Episodic Memory Networks for Learning from Streaming Data
This addresses the problem of inefficient memory usage in lifelong learning for AI systems, though it is an incremental improvement over existing memory-augmented networks.
The paper tackles the scalability issue of memory-augmented neural networks in lifelong learning by proposing Long-term Episodic Memory Networks (LEMN), which use an RNN-based retention agent to selectively replace less important memory entries, achieving significant improvements over rule-based and RL-based baselines on path-finding and question-answering tasks.
Current generation of memory-augmented neural networks has limited scalability as they cannot efficiently process data that are too large to fit in the external memory storage. One example of this is lifelong learning scenario where the model receives unlimited length of data stream as an input which contains vast majority of uninformative entries. We tackle this problem by proposing a memory network fit for long-term lifelong learning scenario, which we refer to as Long-term Episodic Memory Networks (LEMN), that features a RNN-based retention agent that learns to replace less important memory entries based on the retention probability generated on each entry that is learned to identify data instances of generic importance relative to other memory entries, as well as its historical importance. Such learning of retention agent allows our long-term episodic memory network to retain memory entries of generic importance for a given task. We validate our model on a path-finding task as well as synthetic and real question answering tasks, on which our model achieves significant improvements over the memory augmented networks with rule-based memory scheduling as well as an RL-based baseline that does not consider relative or historical importance of the memory.