Merge pull request #153 from webmachinelearning/wchao/fix_bad_links

Fixing bad links and references in the webnn explainer.

explainer.md

@@ -85,7 +85,7 @@ Depending on the underlying hardware capabilities, these platform APIs may make
 
 A core abstraction behind popular neural networks is a computational graph, a directed graph with its nodes corresponding to operations (ops) and input variables. One node's output value is the input to another node. The WebNN API brings this abstraction to the web.
 
-In the WebNN API, the [`Operand`](https://webmachinelearning.github.io/webnn/#operand) objects represent input, output, and constant multi-dimensional arrays known as [tensors](https://mathworld.wolfram.com/Tensor.html). The [`NeuralNetworkContext`](https://webmachinelearning.github.io/webnn/#api-neuralnetworkcontext) defines a set of operations that facilitate the construction and execution of this computational graph. Such operations may be accelerated with dedicated hardware such as the GPUs, CPUs with extensions for deep learning, or dedicated ML accelerators. These operations defined by the WebNN API are required by [models](https://github.com/webmachinelearning/webnn/blob/master/op_compatibility/first_wave_models.md) that address key application use cases. Additionally, the WebNN API provides affordances to builder a computational graph, compile the graph, execute the graph, and integrate the graph with other Web APIs that provide input data to the graph e.g. media APIs for image or video frames and sensor APIs for sensory data.
+In the WebNN API, the [`MLOperand`](https://webmachinelearning.github.io/webnn/#api-mloperand) objects represent input, output, and constant multi-dimensional arrays known as [tensors](https://mathworld.wolfram.com/Tensor.html). The [`MLContext`](https://webmachinelearning.github.io/webnn/#api-mlcontext) defines a set of operations that facilitate the construction and execution of this computational graph. Such operations may be accelerated with dedicated hardware such as the GPUs, CPUs with extensions for deep learning, or dedicated ML accelerators. These operations defined by the WebNN API are required by [models](https://github.com/webmachinelearning/webnn/blob/master/op_compatibility/first_wave_models.md) that address key application use cases. Additionally, the WebNN API provides affordances to builder a computational graph, compile the graph, execute the graph, and integrate the graph with other Web APIs that provide input data to the graph e.g. media APIs for image or video frames and sensor APIs for sensory data.
 
 This [example](https://webmachinelearning.github.io/webnn/#examples) builds, compiles, and executes a graph comprised of three ops, takes four inputs and returns one output.
 
@@ -192,7 +192,7 @@ Neural network operations are mathematical functions. There are about a hundred
 
 In designing the WebNN operations, a proposal to decompose high-level functions to the more rudimentary mathematical operations was considered, with the key benefit of having a reduced number of operations defined. However, such an approach would make the networks more verbose and harder to construct. It'll also risk losing the opportunity to leverage known optimizations for highly reusable functions in the operating systems and in the hardware platforms underneath it. For instance, most operating systems and modern hardware today support widely-used variants of convolutions and recurrent networks out of the box. By decomposing well-known functions into networks of rudimentary mathematical operations, their identities may be lost in the process with opportunities for significant performance gains left behind.
 
-To balance the needs of providing for future extensibility while ensuring maximum reuse and performance optimization opportunity, we chose to include both the standard functions and all the smaller operations making up the functions in the spec. For each high-level function defined, we make sure that all of its decomposed operations are also defined. This way, a newly-conceived function may be represented as a network of our decomposed operations, while a standard function can also be fully supported by the underlying platforms. An elaborate example of this principle is in the way we define the specification of the [gruCell](https://webmachinelearning.github.io/webnn/#api-modelbuilder-grucell) operation as described in its notes.
+To balance the needs of providing for future extensibility while ensuring maximum reuse and performance optimization opportunity, we chose to include both the standard functions and all the smaller operations making up the functions in the spec. For each high-level function defined, we make sure that all of its decomposed operations are also defined. This way, a newly-conceived function may be represented as a network of our decomposed operations, while a standard function can also be fully supported by the underlying platforms. An elaborate example of this principle is in the way we define the specification of the [gruCell](https://webmachinelearning.github.io/webnn/#api-mlgraphbuilder-grucell) operation as described in its notes.
 
 ## Considered alternatives