Looking to the future is always something people have been interested in from the old days. What people can do using their current ability is ambiguous, but this can be achieved psychologically. In this study, a systematic Augmented Reality (AR) system named iCan has been proposed to reveal the ability of what we can do and which way we should choose when facing any junction of choices. iCan has been developed in four steps: basic AR development implementation, building 3D-Models, making the best interaction UX(User-Experience) and test-debug-publish. In the first step, booklet pages have been chosen for AR development instruction. Next, 3D models have made and have been presented. Next, interaction methods are under development. Some individuals have tested the system, and the results show the robustness of the presented system, which there is a hope to change people's future soon.

Many recent machine learning models rely on fine-grained dynamic control flow for training and inference. In particular, models based on recurrent neural networks and on reinforcement learning depend on recurrence relations, data-dependent conditional execution, and other features that call for dynamic control flow. These applications benefit from the ability to make rapid control-flow decisions across a set of computing devices in a distributed system. For performance, scalability, and expressiveness, a machine learning system must support dynamic control flow in distributed and heterogeneous environments.

Batch-splitting (data-parallelism) is the dominant distributed Deep Neural Network (DNN) training strategy, due to its universal applicability and its amenability to Single-Program-Multiple-Data (SPMD) programming. However, batch-splitting suffers from problems including the inability to train very large models (due to memory constraints), high latency, and inefficiency at small batch sizes. All of these can be solved by more general distribution strategies (model-parallelism). Unfortunately, efficient model-parallel algorithms tend to be complicated to discover, describe, and to implement, particularly on large clusters. We introduce Mesh-TensorFlow, a language for specifying a general class of distributed tensor computations. Where data-parallelism can be viewed as splitting tensors and operations along the "batch" dimension, in Mesh-TensorFlow, the user can specify any tensor-dimensions to be split across any dimensions of a multi-dimensional mesh of processors. A Mesh-TensorFlow graph compiles into a SPMD program consisting of parallel operations coupled with collective communication primitives such as Allreduce. We use Mesh-TensorFlow to implement an efficient data-parallel, model-parallel version of the Transformer sequence-to-sequence model. Using TPU meshes of up to 512 cores, we train Transformer models with up to 5 billion parameters, surpassing SOTA results on WMT'14 English-to-French translation task and the one-billion-word Language modeling benchmark. Mesh-Tensorflow is available at https://github.com/tensorflow/mesh

We describe an approach for synthesizing data representations for concurrent programs. Our compiler takes as input a program written using concurrent relations and synthesizes a representation of the relations as sets of cooperating data structures as well as the placement and acquisition of locks to synchronize concurrent access to those data structures. The resulting code is correct by construction: individual relational operations are implemented correctly and the aggregate set of operations is serializable and deadlock free. The relational specification also permits a high-level optimizer to choose the best performing of many possible legal data representations and locking strategies, which we demonstrate with an experiment autotuning a graph benchmark.

We consider the problem of specifying combinations of data structures with complex sharing in a manner that is both declarative and results in provably correct code. In our approach, abstract data types are specified using relational algebra and functional dependencies. We describe a language of decompositions that permit the user to specify different concrete representations for relations, and show that operations on concrete representations soundly implement their relational specification. It is easy to incorporate data representations synthesized by our compiler into existing systems, leading to code that is simpler, correct by construction, and comparable in performance to the code it replaces.

We present an overview of the Saturn program analysis system, including a rationale for three major design decisions: the use of function-at-a-time, or summary-based, analysis, the use of constraints, and the use of a logic programming language to express program analysis algorithms. We argue that the combination of summaries and constraints allows Saturn to achieve both great scalability and great precision, while the use of a logic programming language with constraints allows for succinct, high-level expression of program analyses.

In this paper we present a new approach to modeling finite set domain constraint problems using Reduced Ordered Binary Decision Diagrams (ROBDDs). We show that it is possible to construct an efficient set domain propagator which compactly represents many set domains and set constraints using ROBDDs. We demonstrate that the ROBDD-based approach provides unprecedented flexibility in modeling constraint satisfaction problems, leading to performance improvements. We also show that the ROBDD-based modeling approach can be extended to the modeling of integer and multiset constraint problems in a straightforward manner. Since domain propagation is not always practical, we also show how to incorporate less strict consistency notions into the ROBDD framework, such as set bounds, cardinality bounds and lexicographic bounds consistency. Finally, we present experimental results that demonstrate the ROBDD-based solver performs better than various more conventional constraint solvers on several standard set constraint problems.