Machine learning systems are equipped with artificial intelligence engines that provide these systems with the capability of learning by themselves without having to write programs to do so. They adjust and change programs as a result of being exposed to big data sets. The process of doing so is similar to the data mining concept where the data set is searched for patterns. The difference is in how those patterns are used. Data mining's purpose is to enhance human comprehension and understanding. Machine learning's algorithms purpose is to adjust some program's action without human supervision, learning from past searches and also continuously forward as it's exposed to new data.

The News Feed service in Facebook is an example, automatically personalizing a user's feed from his interaction with his or her friend's posts. The "machine" uses statistical and predictive analysis that identify interaction patterns (skipped, like, read, comment) and uses the results to adjust the News Feed output continuously without human intervention. 

Impact on Existing and Emerging Markets

The NBA is using machine analytics created by a California-based startup to create predictive models that allow coaches to better discern a player's ability. Fed with many seasons of data, the machine can make predictions of a player's abilities. Players can have good days and bad days, get sick or lose motivation, but over time a good player will be good and a bad player can be spotted. By examining big data sets of individual performance over many seasons, the machine develops predictive models that feed into the coach’s decision-making process when faced with certain teams or particular situations. 

General Electric, who has been around for 119 years is spending millions of dollars in artificial intelligence learning systems. Its many years of data from oil exploration and jet engine research is being fed to an IBM-developed system to reduce maintenance costs, optimize performance and anticipate breakdowns.

Over a dozen banks in Europe replaced their human-based statistical modeling processes with machines. The new engines create recommendations for low-profit customers such as retail clients, small and medium-sized companies. The lower-cost, faster results approach allows the bank to create micro-target models for forecasting service cancellations and loan defaults and then how to act under those potential situations. As a result of these new models and inputs into decision making some banks have experienced new product sales increases of 10 percent, lower capital expenses and increased collections by 20 percent. 

Emerging markets and industries

By now we have seen how cell phones and emerging and developing economies go together. This relationship has generated big data sets that hold information about behaviors and mobility patterns. Machine learning examines and analyzes the data to extract information in usage patterns for these new and little understood emergent economies. Both private and public policymakers can use this information to assess technology-based programs proposed by public officials and technology companies can use it to focus on developing personalized services and investment decisions.

Machine learning service providers targeting emerging economies in this example focus on evaluating demographic and socio-economic indicators and its impact on the way people use mobile technologies. The socioeconomic status of an individual or a population can be used to understand its access and expectations on education, housing, health and vital utilities such as water and electricity. Predictive models can then be created around customer's purchasing power and marketing campaigns created to offer new products. Instead of relying exclusively on phone interviews, focus groups or other kinds of person-to-person interactions, auto-learning algorithms can also be applied to the huge amounts of data collected by other entities such as Google and Facebook.

A warning

Traditional industries trying to profit from emerging markets will see a slowdown unless they adapt to new competitive forces unleashed in part by new technologies such as artificial intelligence that offer unprecedented capabilities at a lower entry and support cost than before. But small high-tech based companies are introducing new flexible, adaptable business models more suitable to new high-risk markets. Digital platforms rely on algorithms to host at a low cost and with quality services thousands of small and mid-size enterprises in countries such as China, India, Central America and Asia. These collaborations based on new technologies and tools gives the emerging market enterprises the reach and resources needed to challenge traditional business model companies.

I suspect that many of you are familiar with the term "hard coding a value" whereby the age of an individual or their location is written into the condition (or action) of a business rule (in this case) as shown below:

if customer.age > 21 and customer.city == 'denver'

then ...

Such coding practices are perfectly expectable provided that the conditional values, age and city, never change. They become entirely unacceptable if a need for different values could be anticipated. A classic example of where this practice occurred that caused considerable heartache in the IT industry was the Y2K issue where dates were updated using only the last 2 digits of a four digit number because the first 2 digits were hard-coded to 19 i.e. 1998, 1999. All was well provided that the date did not advance to a time beyond the 1900’s since no one could be certain of what would happen when the millennia arrived (2000). A considerably amount of work (albeit boring) and money, approximately $200 billion, went into revising systems by way of software rewrites and computer chip replacements in order to thwart any detrimental outcomes. It is obvious how a simple change or an assumption can have sweeping consequences.

You may wonder what Y2K has to do with Business Rule Management Systems (BRMS). Well, what if we considered rules themselves to be hard-coded. If we were to write 100s of rules in Java, .NET or whatever language that only worked for a given scenario or assumption, would that not constitute hard-coded logic? By hard-coded, we obviously mean compiled. For example, if a credit card company has a variety of bonus campaigns, each with their own unique list of rules that may change within a week’s time, what would be the most effective way of writing software to deal with these responsibilities?

Java still has its place in the world of software development, but is it quickly becoming obsolete by the more dynamically enabled Python programming language? The issue is hotly contested by both sides of the debate. Java experts point out that Java is still being developed with more programmer friendly updates. Python users swear that Java can take up to ten times longer to develop. Managers that need to make the best decision for a company need concrete information so that an informed and rational decision can be made.

First, Java is a static typed language while Python is dynamically typed. Static typed languages require that each variable name must be tied to both a type and an object. Dynamically typed languages only require that a variable name only gets bound to an object. Immediately, this puts Python ahead of the game in terms of productivity since a static typed language requires several elements and can make errors in coding more likely.

Python uses a concise language while Java uses verbose language. Concise language, as the name suggests, gets straight to the point without extra words. Removing additional syntax can greatly reduce the amount of time required to program.  A simple call in Java, such as the ever notorious "Hello, World" requires three several lines of coding while Python requires a single sentence. Java requires the use of checked exceptions. If the exceptions are not caught or thrown out then the code fails to compile. In terms of language, Python certainly has surpassed Java in terms of brevity.

Additionally, while Java's string handling capabilities have improved they haven't yet matched the sophistication of Python's. Web applications rely upon fast load times and extraneous code can increase user wait time. Python optimizes code in ways that Java doesn't, and this can make Python a more efficient language. However, Java does run faster than Python and this can be a significant advantage for programmers using Java. When you factor in the need for a compiler for Java applications the speed factor cancels itself out leaving Python and Java at an impasse.

While a programmer will continue to argue for the language that makes it easiest based on the programmer's current level of knowledge, new software compiled with Python takes less time and provides a simplified coding language that reduces the chance for errors. When things go right, Java works well and there are no problems. However, when errors get introduced into the code, it can become extremely time consuming to locate and correct those errors. Python generally uses less code to begin with and makes it easier and more efficient to work with.

Ultimately, both languages have their own strengths and weaknesses. For creating simple applications, Python provides a simpler and more effective application. Larger applications can benefit from Java and the verbosity of the code actually makes it more compatible with future versions. Python code has been known to break with new releases. Ultimately, Python works best as a type of connecting language to conduct quick and dirty work that would be too intensive when using Java alone. In this sense, Java is a low-level implementation language. While both languages are continuing to develop, it's unlikely that one language will surpass the other for all programming needs in the near future.

In programming, memory leaks are a common issue, and it occurs when a computer uses memory but does not give it back to the operating system. Experienced programmers have the ability to diagnose a leak based on the symptoms. Some believe every undesired increase in memory usage is a memory leak, but this is not an accurate representation of a leak. Certain leaks only run for a short time and are virtually undetectable.

Memory Leak Consequences

Applications that suffer severe memory leaks will eventually exceed the memory resulting in a severe slowdown or a termination of the application.

How to Protect Code from Memory Leaks?

Preventing memory leaks in the first place is more convenient than trying to locate the leak later. To do this, you can use defensive programming techniques such as smart pointers for C++.  A smart pointer is safer than a raw pointer because it provides augmented behavior that raw pointers do not have. This includes garbage collection and checking for nulls.

If you are going to use a raw pointer, avoid operations that are dangerous for specific contexts. This means pointer arithmetic and pointer copying. Smart pointers use a reference count for the object being referred to. Once the reference count reaches zero, the excess goes into garbage collection. The most commonly used smart pointer is shared_ptr from the TR1 extensions of the C++ standard library.

Static Analysis

The second approach to memory leaks is referred to as static analysis and attempts to detect errors in your source-code. CodeSonar is one of the effective tools for detection. It provides checkers for the Power of Ten coding rules, and it is especially competent at procedural analysis. However, some might find it lagging for bigger code bases.

How to Handle a Memory Leak

For some memory leaks, the only solution is to read through the code to find and correct the error. Another one of the common approaches to C++ is to use RAII, which an acronym for Resource Acquisition Is Initialization. This approach means associating scoped objects using the acquired resources, which automatically releases the resources when the objects are no longer within scope. RAII has the advantage of knowing when objects exist and when they do not. This gives it a distinct advantage over garbage collection. Regardless, RAII is not always recommended because some situations require ordinary pointers to manage raw memory and increase performance. Use it with caution.

The Most Serious Leaks

Urgency of a leak depends on the situation, and where the leak has occurred in the operating system. Additionally, it becomes more urgent if the leak occurs where the memory is limited such as in embedded systems and portable devices.

To protect code from memory leaks, people have to stay vigilant and avoid codes that could result in a leak. Memory leaks continue until someone turns the system off, which makes the memory available again, but the slow process of a leak can eventually prejudice a machine that normally runs correctly.

Related:

The Five Principles of Performance

In Demand IT Skills