Fix spelling

Contributing.md

@@ -4,7 +4,7 @@ Adding new features, improving documentation, fixing bugs, writing new tests, de
 
 We kindly request that any bug fixes be initiated through GitHub pull requests (PRs). 
 
-## Prerequisits
+## Prerequisites
 If you are new to GitHub, you can find useful documentation [here][1], and information on the `git` version control system in this [handbook][2].
 
 ## Coding convention

Credit Card Fraud/CreditCardInferenceViewController.mm

@@ -51,7 +51,7 @@
 void createContexts()
 {
 
-  cout << "Initalizing HElib . . ." << endl;
+  cout << "Initializing HElib . . ." << endl;
 
   shared_ptr<HeContext> hePtr;
 

Credit Card Fraud/README.md

@@ -5,11 +5,11 @@ This FHE toolkit example demonstrates a use case in the finance domain as well a
 
 ## Relation to a real use case
 
-The concept of providing fully outsourced, but fully encrypted computation to a cloud provider is a major motivating factor in the feild of FHE. This use case example shows the capability of the toolklit to build such applications. While the client and server are not literally separated (nor demonstrating true remote cloud computation), the concepts generalize.  
+The concept of providing fully outsourced, but fully encrypted computation to a cloud provider is a major motivating factor in the field of FHE. This use case example shows the capability of the toolklit to build such applications. While the client and server are not literally separated (nor demonstrating true remote cloud computation), the concepts generalize.  
 
 With respect to the realism of the use-case, the network we use is based on a network architecture implemented by the Kaggle community (in [Credit Card Fraud Detection using Neural Networks](https://www.kaggle.com/omkarsabnis/credit-card-fraud-detection-using-neural-networks)). The example network used for the sample application was trained using a realistic dataset taken from [Credit Card Fraud Detection](https://www.kaggle.com/mlg-ulb/creditcardfraud?select=creditcard.csv). This dataset contains actual anonymized transactions made by credit card holders from September 2013, and the data set containes labeled about transactions being fraudulent or genuine.
 
-Since this is an educational example, to demonstrate one class of use case possibly with the toolkit, we have opted to  simplify the architecture to demonstrate fast runtime with low memory use. In particular, this is achieved by replacing the activation functions with a square function, which reduced precision from 0.83 to 0.8. We encourage our users to explore the full power of the included example FHE AI library that includes more advanced tools that allow maintaining higher precision as well as better performance. Specifically, if you would like to experiment with the trade off between increasing security and model resources, you can increase security by adjusting the 'm' parameter (line 72) to a larger power of 2. By setting the value to 2^17, the sample applicaiton would reach a security level that is equivalent to a 287 bits symmetric key. Doing so will increase security but will also increase CPU time, memory usage, and I/O time. The full version of this library uses advanced techniques to keep all of these measures low even for higher levels of security, as well as, allow parallelization of the process.
+Since this is an educational example, to demonstrate one class of use case possibly with the toolkit, we have opted to  simplify the architecture to demonstrate fast runtime with low memory use. In particular, this is achieved by replacing the activation functions with a square function, which reduced precision from 0.83 to 0.8. We encourage our users to explore the full power of the included example FHE AI library that includes more advanced tools that allow maintaining higher precision as well as better performance. Specifically, if you would like to experiment with the trade off between increasing security and model resources, you can increase security by adjusting the 'm' parameter (line 72) to a larger power of 2. By setting the value to 2^17, the sample application would reach a security level that is equivalent to a 287 bits symmetric key. Doing so will increase security but will also increase CPU time, memory usage, and I/O time. The full version of this library uses advanced techniques to keep all of these measures low even for higher levels of security, as well as, allow parallelization of the process.
 
 Lastly, with respect to realism and performance, the sample application is designed to processes data in batches of m/4 (the 'm' is the same variable that was specified at line 72 as mentioned previously). E.g., if the demo as provided is configured to use m=2^15, the software will process batches of 8,192 transactions. Working in batches usually provides high throughput but low latency, and are inefficient when there are not enough samples to fill the required batch size. The full version of this library uses advanced techniques to allow efficient, low latency inference for any batch size, irrespective of m.