Deep Learning vs. Machine Learning

Deep Learning is a category of machine learning with special advantages for some tasks and disadvantages for others.

Machine learning workflows begin by identifying features within data sets. For structured information with relatively few columns and rows, this is straightforward. Most practical business predictions such as classification and regression fall into this category.

Unstructured data, such as image and voice, have vast numbers of “features” in the form of individual pixels or wave forms. Identifying those features to structured AI algorithms is tedious or impossible. Deep Learning is a technique where the AI algorithm itself extracts progressively higher levels of feature recognition, passing information through potentially hundreds of neural network layers. Deep learning algorithms power image and speech recognition for driverless cars and hand-free speakers

Plusses of Deep Learning

  • Scale – Deep learners can handle vast amounts of data, and they always improve with more data. Shallow learning converges and stops improving with additional data.
  • Dimensions – Deep learners can move past the limitations of a few hundred columns to perform well on very wide structured data set.
  • Non-Numeric – Deep learning brings AI into the human realm of speech and vision, which serve people in new and valuable ways.

Minuses of Deep Learning

  • Training Data – Deep learners need labeled data from which to learn. Amassing sufficient examples for recognition accuracy to be learned can be daunting.
  •  Not for Small Data Sets – Data sets that are too simple or too small cause deep learners to fail by overfitting
  • Resource Consumption – Deep learning on vast data stores can require days or weeks of processing on a single problem. 

The take-away: Deep learners are great for unstructured data and may be useful for classification with large and detailed structured data sets. Squark includes deep learners in the stack of algorithms it uses for AutoML. You will know from the Squark Leaderboard whether deep learning was a winner.

Classification Types and Uses

Classifications are the most frequently used—and most useful—prediction types.

Classifications are predictions that separate data into groups. Binary Classification produces “yes-no” or “in-out” answers when there are only two choices. Multi-Class Classification applies when there are three or more possibilities and shows probabilities for each.

Binary Examples

  • Churn – Which customers are in danger of leaving? If you knew, you could implement targeted retention tactics such as special offers or customer service outreach.
  • Conversion – Which prospects are most likely to be ready to move to the next stem in the buying cycle? Knowing means focusing sales resources on the best leads.
  • Risk – Which populations are likely to experience negative outcomes? Understanding helps guide actions to mitigate risks.

Multi-Class Examples

  • Cross-Sell/Up-Sell – Which customers are most likely to buy which additional products or services? Targeting them with the right offers lifts sales with high efficiency.
  • Personalization – Which content will resonate with which person? Optimizing websites, social media, and email is easy when you know.
  • Ad Targeting – Which prospects are most likely to respond to your multiplicity of ads and media? Spending is more effective when you know your audiences.

The take-away: Classifications are among the most accessible and highest-return prediction types for AutoML. Predictions on each row include not only the classes, but the probabilities associated with each. Think of a burning classification question and Squark can help you begin predicting right away.

Operationalizing AutoML: ML Ops

Predictions are interesting on their own. They are valuable when put into production.

Operationalizing AutoML – often called “ML Ops” – means putting AutoML predictions into regular workflows to change business outcomes. Here are a few ways do it.

Graphical Interface
Saas AutoML tools have a GUI that enables training data ingestion, data prep, feature engineering, and model building. Once an optimal model is made, predictions are created on production data. Simply exporting the predictions from the GUI delivers a data file for people or other systems to act upon. For example, predictions the prioritize sales lead follow-up could be handed to sales ops as a daily call list, or could be imported to a marketing automation system as an email segmentation list.

API
Application Programming Interfaces (APIs) are specifications that describe how dissimilar systems can communicate reliably. AutoML APIs can be hooked to enterprise systems to automate export of predictions to eliminate manual file handling.

Model Export
AutoML produces executable code, typically Java bytecode, that can be run wholly outside the AutoML tool itself. Models so exported can be run again and again on production data as often as required. When models are improved based on new training data, ML ops can simply replace the executable code with the new models.

Custom Deployment
Some applications, such as real-time predictions, require close integration with cooperating systems. Customized data pipelines can be created to manage these processes.

The take-away: You can begin using AutoML results to improve business performance right now. As needs expand, there are many options for blending ML Ops into automated workflows, and Squark can help with all of them.

AI Changed Dramatically in Only 9 Months

Productive uses for AI are closer at hand than ever due to the rise of AutoML.

Beginning in 2019, advancements in AI have replaced obstacles with tools to benefit from its power. Foremost among them is Automated Machine Learning (AutoML), which does not require programming or scripting of any kind. Here are some examples.

Business Analysts
“AI is the new BI” is a theme repeated for years by journalists and product marketers. AI is now simple enough that business analysts can make reliable predictions without being data scientists or programmers. Moving from visualizing trends with BI to predicting the future—record-by-record—with AI is transforming the way businesses run. AutoML is what makes that possible.

Citizen Data Scientists
Researchers and designers who need to understand patterns know their problems and data very well, but not necessarily the AI algorithms that can extract information they need. AutoML abstracts algorithm selection and use to produce results quickly.

Data Science Professionals
AutoML does not replace the need for true data scientists, whose expertise in solving complicated AI problems cannot be replicated. Nevertheless, AI pros use AutoML to glean insights on problems to make custom work more efficient. In addition, offloading simpler BI problems to AutoML keeps them focused where they are most needed.

Conclusion: AutoML makes achieving the benefits of AI simpler for everyone.

Why AI for Marketing and Sales?

Follow the money to see why marketing and sales are the most common applications for AI.

Instant Payback
Small improvements in marketing and sales can produce large returns quickly. Think of the impact of gaining a few percentage points on lead conversions, forecast accuracy, content targeting, and ad performance. Knowing which customers will buy, what they will buy, and when they will buy delivers value on both revenue and cost sides of the ledger.

Plenty of Data
More information than ever is available in CRM, marketing automation, and customer data platforms. AI—in the form of Automated Machine Learning (AutoML)—is really good at finding patterns in all that data to predict the future.

Easy
AutoML does not require programming or formula creation in order to make accurate predictions. Models can be made and refined rapidly. This is particularly important in supporting nimble marketing and sales processes.

AutoML insights for marketing and sales are easy to monetize and straightforward to execute. That makes them great places to amplify the benefits of AI.

Machine Learning vs. Statistics

Statistics and machine learning differ in method and purpose. Which is superior depends upon your goals.

Statistics is a subset of mathematics that interprets relationships among variables in data sets. Statisticians make inferences and estimate values based solely on data collected during a specific period, a rearward-looking view. Understanding how data was collected and the distributions of populations must be considered in model building. Statistics are useful where assumptions and probabilities must be mathematically auditable, such as when publishing a scientific paper on experimental observations.

Machine Learning (specifically, supervised learning) is a subset of computer science that uses past data to predict the future. The forward-looking view relies on training models using data sets of known outcomes and testing accuracy against test sets sequestered from the training data. The hold-back process proves that predictions on future data will be similarly accurate. Machine learning excels when there are large numbers of variables and records in data sets.

Conclusion: Use statistics for “court of law” explanations of what happened in the past. Use machine learning to make record-by-record predictions of future outcomes.

How Important Is Explainability?

How Important Is Explainability?

Understanding the inner workings of ML algorithms may distract from realizing benefits from good predictions.

Explainability Explained

With the rise of artificial intelligence has come skepticism. Mysteries of how AI works make questioning the “black box” natural. “Explainability” refers to being able to trace and follow the logic AI algorithms use to form their conclusions. In some cases—particularly in unsupervised learning—the answer is, “We don’t know.” How disconcerting is that? Whether or not the answers are valid, not being able to “show your work” engenders suspicions.

Supervised learning is different. Algorithms such as trees and linear regressions have clearly defined math that humans could follow and arrive at the same answers as automated machine learning (AutoML), if only they had time to work through millions of calculations. Nevertheless, unfamiliarity with the data science of supervised learning also causes doubts.

Explainability could be a concern when …

  • Making life-or-death decisions
  • Proving court-of-law-style certainty
  • Completely eradicating biases

Fortunately, most practical business decisions do not need to pass those tests.

How to Convince Decision Makers to Trust AutoML

First of all, emphasize the ROI of good predictions. In marketing, one of the most common use cases for AutoML, predictions only have to be a little better than a coin flip to deliver huge returns. Next, show the evidence:

  • Model accuracy is calculated by comparing AutoML predictions to known results. This same accuracy will prevail for new, unseen data.
  • Squark shows lists of the factors that were most predictive, which explains enough of the model’s logic to inspire confidence.
  • Data features can easily be added and removed to test biases and understand predictive behaviors.

A Useful Analogy

Before using Google Maps to find the best route to your destination, do you investigate which AI algorithms it used and why it chose that exact route? Of course not. The reasons are simple:

  • The algorithms are not understandable unless you are a data scientist.
  • The results are usually very good.
  • Time wasted checking the process delays reaching your goal.

Squark displays algorithm performance information along with prediction results. This shows how a multiplicity of models were evaluated to make sure the best one was utilized. Think of the lower-ranked models as the gray, “12 minutes slower” routes and start your journey with confidence.

How are Supervised and Unsupervised Learning Different?

Showing the way vs. stumbling in the dark – there are applications for both.

Supervised

Supervised Learning shows AutoML algorithms sets of known outcomes from which to learn. Think of classroom drills, or giving a bloodhound the scent.

Supervised learning relies on training data containing labels for data columns (features). Known outcomes must be included for the columns to be predicted on fresh data.

Use Supervised Learning for…

  • Performance-based predictions
  • Scoring the likelihood of things to happen
  • Forecasting outcomes

Classifications where there are two (binary) or more (multi-class) possibilities are use cases for supervised learning . Regressions—predicting scalar numerical values such as forecasts—are also suited to supervised learning.

Unsupervised

Unsupervised learning happens when AutoML algorithms are given unlabeled training data and must make sense of it without any instruction. Such machines “teach themselves” what result to produce.

Unsupervised learning algorithms look for structure in the training data, like finding which examples are similar to each other and grouping them into clusters.

Use Unsupervised Learning for…

  • Understanding co-occurrence
  • Detecting hidden data relationships
  • Extracting data

Clustering, market basket analyses, and anomaly detection are common use cases for unsupervised learning.

What Is A Confusion Matrix?

What Is A Confusion Matrix?

The most aptly named AI term is actually simple.

A Confusion Matrix is a table that shows how often an AI classifier gets confused predicting true and false conditions. Here is a simple example of a Confusion Matrix for a model that classifies whether a fruit is an orange or not, out of a sample of 166.

How Well Did My Model Do?

As you can see from the table, the classifier was pretty accurate overall. It was correct (true positives and true negatives) 155 times, or 93.37% of the time. It did very well at predicting when fruits were oranges – only one wrong (false negative), or about 99%. It was not as good at predicting when they were not oranges – 83.3% right and 16.7 % wrong (false positives).

Confusion matrices are especially informative when considering the consequences of false negatives versus false positives in your use cases.

What Is Overfitting and How Do I Avoid It?

What Is Overfitting?

Telltale Super-Accuracy on Training Data

When machine learning models show exceptional accuracy on training data sets, but perform poorly on new, unseen data, they are guilty of overfitting. Overfitting happens when models “learn” from noise in data instead of from true signal patterns.

How to Avoid Overfitting

Detecting overfitting is the first step. Comparing accuracy against a portion of training that was data set aside for testing will reveal when models are overfitting. Techniques to minimize overfitting include:

  • Tuning Hyperparameters – Hyperparameters are descriptions of data set properties—information about the data, not the data itself. Hyperparameters can be used to adjust settings for different families of machine learning algorithms so they perform well and do not overfit.
  • Cross-Validation – Cross-validation splits training data into additional train-test sets to tune hyperparameters iteratively, without disturbing the initial test set-aside data.
  • Early Stopping – Machine learning algorithm training generally improves model performance with more attempts—up to a point. Comparing model performance at each building iteration and stopping when accuracy no longer improves prevents overfitting.

Squark Seer automatically employs these and other approaches to minimize overfitting. As always, get in touch if you have questions about Overfitting or any other Machine Learning topic. We’re happy to help.