Humalytica Strikes Again

neural_networkingThe Analytics Post, Wednesday, August 6, 2015

The most unpopular, fastest decaying piece of analytics software today, Humalytica, a.k.a. the human mind, is reportedly trying to engage model builders to use rational thought when selecting an algorithm. Ted Neuron of Auto Neural Networks, Inc. made this comment upon hearing the news: “Our algorithm is extremely powerful, can model even very complex relationships, and there is no need to understand the underlying data. Humalytica requires too much thought and is prone to selecting the best approach, rather than the easiest one.”

Dr. Jeffrey Strickland, representing Humalytica, was asked to respond to Mr. Neuron’s statement. “Ted is correct. However what he did not say is his algorithm is prone to overfitting, incurs a long training time, requires significant computing power for large datasets, and the model is essentially unreadable. There are times when it is the best algorithm for the problem at hand, but not always.”

Lonnie Matrix from Support Vector Machines, Inc., was asked for her opinion. “It is simply a fact that SVMs can model complex, nonlinear relationships and is robust to noise (because they maximize margins). One cannot say this of Humalytica or ANNs. Humalytica requires too much thought. You can model any problem with SVMs and get an excellent answer.”

Again, Dr. Strickland was asked to respond. “Lonnie is correct. However, what she did not say is in order to use SVMs, you need to select a good kernel function, which may require the use of Humalytica. Moreover, with SVMs model parameters are difficult to interpret, they sometimes encounter numerical stability problems, and they require significant memory and processing power. Having said that, SVMs do have their place. They are good for text classification, image classification and handwriting recognition.”

Though not asked, Mr. Rogers, of Mr. Rogers’s Neighborhood, had this to say regarding his own K-Nearest Neighbors algorithm. “My algorithm is simple, powerful, requires no training involved (“lazy”), and naturally handles multiclass classification and regression. No need for Humalytica here. Just use KNNs for all your problems. By the way, would you be my neighbor?”

We asked Dr. Strickland to respond. “Mr. Rogers is correct. However, KNNs are expensive and slow to predict new instances, one must define a meaningful distance function, and they perform poorly on high-dimensionality datasets. They are good for low-dimensional datasets, computer security, like intrusion detection, crime detection and fault detection in semiconductor manufacturing.”

Having heard several sides of the issue the Analytics Post turned to the more traditional Alice Stats, from Linear Regression, Inc. She had this to say: “Linear regression is the obvious choice for any problem. It is grounded in probability and statistics. It runs very fast (runs in constant time), it is easy to understand the model and it is less prone to overfitting. No need for Humalytica hear.”

Dr. Strickland was asked to respond. “Alice is correct. However, she failed to point out that linear regression is unable to model complex relationship and it is unable to capture nonlinear relationships without first transforming the inputs. It is good for the first look at a dataset and numerical data with lots of features.”

Although the debate is still in progress, can Humalytica compete with these growing, popular giants? This reporter believes that the human mind is due a good and long vacation.

Profile_PicAuthored by:
Jeffrey Strickland, Ph.D.

Jeffrey Strickland, Ph.D., is the Author of Predictive Analytics Using R and a Senior Analytics Scientist with Clarity Solution Group. He has performed predictive modeling, simulation and analysis for the Department of Defense, NASA, the Missile Defense Agency, and the Financial and Insurance Industries for over 20 years. Jeff is a Certified Modeling and Simulation professional (CMSP) and an Associate Systems Engineering Professional (ASEP). He has published nearly 200 blogs on LinkedIn, is also a frequently invited guest speaker and the author of 20 books including:

  • Operations Research using Open-Source Tools
  • Discrete Event simulation using ExtendSim
  • Crime Analysis and Mapping
  • Missile Flight Simulation
  • Mathematical Modeling of Warfare and Combat Phenomenon
  • Predictive Modeling and Analytics
  • Using Math to Defeat the Enemy
  • Verification and Validation for Modeling and Simulation
  • Simulation Conceptual Modeling
  • System Engineering Process and Practices

Connect with Jeffrey Strickland
Contact Jeffrey Strickland

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s