Learning machine learning

29 December 2017

Learning machine learning

Machine Learning is a type of Artificial Intelligence that provides computers the ability to learn or to improve without being explicitly programmed.

Instead of writing a brand new algorithm, machine learning tools enable systems to develop and refine algorithms, by finding patterns in huge amounts of data. It focuses on the development of computer programs that can access data and use it to learn for themselves.

Machine learning methods include observation of data and a set of instructions, which helps systems to learn without interaction or assistance.

Types of Machine Learning

1. Supervised Learning:

The models which use past data for observation or processing output are called supervised models. This type of model is already a trained model with a training dataset, they can present the output by processing the training dataset. In general, they are used for the classification of objects, categories etc.
Example: A machine learning model for the classification of vehicles can be supervised models as a dataset for the classification of vehicles category can be found easily and it will not change in decades.

2. Unsupervised Learning:

When sufficient data is not available in the training dataset or processing output requires more human interaction such as observing habits, interests and activities, the model improves the prediction over time by learning all those things.
Example: A machine learning model predicts a more accurate combination of products to customers buying either product.

3. Reinforcement Learning:

It is more like an unsupervised learning model, the major difference between both of them is Reinforcement learning model can present more accurate results after continuous processing of input data with less human interaction/assistance.
Example: A machine learning model to predict the profit/loss of some organization or someone’s income or savings by a few input parameters.

4. Applications of Machine Learning:

1. Banking & Financial Services:

Bankers can use machine learning to identify customers based on their track record to set credit limits, grant loans, avail rewards, and for other services like preventing frauds, investment trends and trading prediction.

2. Healthcare:

Doctors can use patients’ symptoms to compare with other patients’ already recorded symptoms to identify disease and suitable treatment.

3. Retail:

Retailers use machine learning for the classification of most selling products and less selling products, selling trends by brands and a combination of products to suggest more products that result in more selling, also it is used for customer loyalty programs.

4. Oil and Gas:

Engineers use it to find new sources of energy and minerals. It is also used to detect/predict sensor failure and streamline oil distribution to make it more efficient and cost-effective.

5. Government:

They use multiple sources of data, to be mined and get insights for more efficiency and to detect frauds. They also use it to maintain an equal ratio between the supply and demand of public safety and utilities.

6. Marketing and Sales:

eCommerce sites use ML with your buying history, wishlist and watched items and suggest users more relevant/personalized items and offers to users.

7. Transportation:

On-demand cab services use ML to identify rush in the city and loyalty points for users and drivers. Real-time traffic maps predict better routes with real-time updates based on insights.

Machine Learning and Apple

Apple has a long history with machine learning. Apple has been using machine learning with its products for a decade or more. E.g. next word prediction on a keyboard, face detection on Photos, etc.

Apple unveiled NSLinguisticTagger with iOS 5 for Natural Language Processing (NLP). For better performance and low-level access, Metal was introduced then they brought Basic Neural Network Subroutines (BNNS) into the Accelerate framework.

In 2017, Apple introduced a new framework called Core ML to integrate already trained models (supervised models) in the app along with Vision. It allows various ranges of models for integration as it uses Accelerate and Metal to optimize the performance of Core ML. It enables users to process the data on the device itself without worrying about leaving the data to be analyzed.

To incorporate AI in the app, Developers don’t need to be experts in AI or ML to deliver an experience powered by AI and ML within their app. According to Apple, they will take care of the technical side of incorporating ML, which allows developers to focus only on building user experiences.

Apple also listed a few domains that can be used with apps: