Abstract:Deep neural network (DNN) and its variants have been extensively used for a wide spectrum of real applications such as image classification, face/speech recognition, fraud detection, and so on. In addition to many important machine learning tasks, as artificial networks emulating the way brain cells function, DNNs also show the capability of storing non-linear relationships between input and output data, which exhibits the potential of storing data via DNNs. We envision a new paradigm of data storage, "DNN-as-a-Database", where data are encoded in well-trained machine learning models. Compared with conventional data storage that directly records data in raw formats, learning-based structures (e.g., DNN) can implicitly encode data pairs of inputs and outputs and compute/materialize actual output data of different resolutions only if input data are provided. This new paradigm can greatly enhance the data security by allowing flexible data privacy settings on different levels, achieve low space consumption and fast computation with the acceleration of new hardware (e.g., Diffractive Neural Network and AI chips), and can be generalized to distributed DNN-based storage/computing. In this paper, we propose this novel concept of learning-based data storage, which utilizes a learning structure called learning-based memory unit (LMU), to store, organize, and retrieve data. As a case study, we use DNNs as the engine in the LMU, and study the data capacity and accuracy of the DNN-based data storage. Our preliminary experimental results show the feasibility of the learning-based data storage by achieving high (100%) accuracy of the DNN storage. We explore and design effective solutions to utilize the DNN-based data storage to manage and query relational tables. We discuss how to generalize our solutions to other data types (e.g., graphs) and environments such as distributed DNN storage/computing.

DeepMapping: Learned Data Mapping for Lossless Compression and Efficient Lookup

3D LiDAR Map Compression Using Deep Neural Network.

Decomposition of Neural Discrete Representations for Large-Scale 3D Mapping

Learning-Based Data Storage [Vision] (Technical Report)

Deep Model Transferability from Attribution Maps

Deep Memory-Augmented Proximal Unrolling Network for Compressive Sensing

A Weight Mapping Strategy for More Fully Exploiting Data in CIM-Based CNN Accelerator

DeepPointMap: Advancing LiDAR SLAM with Unified Neural Descriptors

SemiMap: A Semi-Folded Convolution Mapping for Speed-Overhead Balance on Crossbars.

DMVC: Deep Mesh Vertex Compression

Activation Map Compression through Tensor Decomposition for Deep Learning

DeepMIF: Deep Monotonic Implicit Fields for Large-Scale LiDAR 3D Mapping

Optimizing for In-memory Deep Learning with Emerging Memory Technology

Memory and Computation Coordinated Mapping of DNNs Onto Complex Heterogeneous SoC.

Dictionary Pair-based Data-Free Fast Deep Neural Network Compression

A Spatial Mapping Algorithm with Applications in Deep Learning-Based Structure Classification

Deep Learning with Passive Optical Nonlinear Mapping

A Broad-Spectrum and High-Throughput Compression Engine for Neural Network Processors

Towards Large-Scale Incremental Dense Mapping using Robot-centric Implicit Neural Representation

Mind Mappings: Enabling Efficient Algorithm-Accelerator Mapping Space Search

Memory-Scalable and Simplified Functional Map Learning