AI: A Taxonomy of Machine Learning and Deep Learning Algorithms

Once again an excellent post by Ajit Jaokar:   Thanks Ajit!

Below is just the intro overview, the much longer post comes through when you click through to Linkedin.   Nicely done, incudes as part of the taxonomy a number of typical usage descriptions.

Artificial Intelligence #5 : A taxonomy of machine learning and deep learning algorithms

Published on May 25, 2021   By Ajit Jaokar

Course Director: Artificial Intelligence: Cloud and Edge Implementations - University of Oxford

Like the Glossary I posted last week, there is no taxonomy for machine learning and deep learning algorithms.

Most ML/DL problems are classification problems, and a small subset of algorithms can be used to solve most of them (ex: SVM. Logistic regression or Xgboost). In that sense, a full taxonomy maybe an overkill. However, if you really want to understand something, you need to know acquire knowledge of a repertoire of algorithms – to overcome the known unknowns problem.

In this post, rather than present a taxonomy, I present a range of taxonomy approaches for machine learning and deep learning algorithms. Some of these are mathematical. If you are just beginning data science, start from the non-mathematical approaches to taxonomy. Don't be tempted to go for the maths approach. But if you have an aptitude towards maths, you should consider the maths approach because it gives you a deeper understanding. Also, I am a bit biased because many in my network in Oxford, MIT, Cambridge, Technion etc would also take a similar maths-based approach.

Finally, I suggest one specific approach to taxonomy which I like and find most complete. It is complex but it is free to download.

Taxonomy approaches

Firstly, the approach from Jason Brownlee is always a good place to start because its pragmatic and implementable in code in A tour of machine learning algorithms. Note that these are machine learning algorithms (not deep learning algorithms). A more visual approach is below source packt.  .... " 

Sometimes Deep Learning Needs Help

Have had a long time interest in language, and as to how and why AI/DeepLearning, can serve in practice  a means of (mostly) recognizing via an learned taxonomy, complex structures like birds, trees or mushrooms.   In Penn's Language Laboratory blog there is a post about this use of taxonomy and parallel distributed processing to provide recognizing 'intelligence'.   But yet more telling and humorous is the first comment, which makes a point that recognizing intelligence can, depending on context, need some serious help.  Serves a good caution to us today.

Towards a Taxonomy for Automated Assistants

Like the idea of identifying and constructing tasks for assistants so they can be more readily be challenged and compared.   This article suggests this be done and gives some examples.

A Taxonomy of Automated Assistants
By Jerrold M. Grochow
Communications of the ACM, April 2020, Vol. 63 No. 4, Pages 39-41  10.1145/3382746

Automated cars are in our future—and starting to be in our present. In 2014, the Society of Automotive Engineers (SAE) published the first version of a taxonomy for degree of automation in vehicles from Level 0 (not automated) to Level 5 (fully automated, no human intervention necessary).8 Since then, this taxonomy has gained wide acceptance—to the point where everyone from the U.S. government (used by the NHTSA5) to auto manufacturers to the popular press are talking in terms of "skipping level 3" or "everyone wants a level 5 car."1 As technology gets developed and improved, having an accepted taxonomy helps ensure people can talk to each other and know they are talking about the same thing. It is time for one of our computing organizations (perhaps ACM?) to develop an analogous taxonomy for automated assistants. With Siri, Alexa, Cortana, and cohorts selling in the "tens of millions"2 and with more than 20 competitors on the market,7 having an easily understandable taxonomy will help practitioners and end users alike.

There is already a significant body of literature aimed at improving the design and use of automated assistants in both industry and academic arenas (with a variety of category names for these devices and systems, using some combination of "automated," "digital," "smart," "intelligent," "personal," "agent," and "assistant"), as the bibliographies of cited works show. Some recent work focused on task content, use cases, and features. The task content of human activity has been widely studied over a long period of time, but Trippas et al.9 note that "how intelligent assistants are used in a workplace setting is less studied and not very well understood." While not presenting a taxonomy of assistants, this type of task content analysis could be used as an aid in intelligent assistant design. Similarly, Mehrotra et al.4 studied interaction with a desktop-based digital assistant with an eye to "help guide development of future user support systems and improve evaluations of current assistants." Knote et al.3 evaluated 115 "smart personal assistants" by literature and website review to create a taxonomy based on cluster analysis of design characteristics such as communications mode, direction of interaction, adaptivity, and embodiment (virtual character, voice), and so forth—a technology and features-based taxonomy. A commercial study of 22 popular "intelligent ... or automated personal assistants"7 reported "Intelligent Agents can be classified based on their degree of perceived intelligence and capability such as simple reflex agents, model-based reflex agents, goal-based agents, utility-based agents and learning agents." While this is an arguably useful taxonomy, it also primarily addresses the technology used and not the actual use of the automated assistant. The website additionally presents editor and user ratings of ease of use, features, and performance that may be of value to end users. .... '

Defining Elements of Knowledge Management

Good non-technical piece in Forbes on elements of knowledge management that are addressed by taxonomies and ontologies.

Taxonomies, Ontologies and Machine Learning: The Future of Knowledge Management By Kurt Cagle in Cognitive World

Taxonomies, Ontologies and Machine Learning: The Future of Knowledge Management
"Taxonomies classify, ontologies specify"   ... 

Identifying Plant Species

Another example of using many images to train via AI.   Also describes the data needs for such a process, done via Deep Learning neural methods.

From the CACM: 
Digitizing plant specimens is opening up a whole new world for researchers looking to mine collections from around the world.

Computer algorithms trained on the images of thousands of preserved plants have learned to automatically identify species that have been pressed, dried and mounted on herbarium sheets, researchers report. ....  " 
Artificial Intelligence Identifies Plant Species for Science  In Nature 
" .... Bonnet's team had already made progress automating plant identification through the Pl@ntNet project. It has accumulated millions of images of fresh plants — typically taken in the field by people using its smartphone app to identify specimens.

Researchers trained similar algorithms on more than 260,000 scans of herbarium sheets, encompassing more than 1,000 species. The computer program eventually identified species with nearly 80% accuracy: the correct answer was within the algorithms’ top 5 picks 90% of the time. That, says Wilf, probably out-performs a human taxonomist by quite a bit. .... " 

Service Design

Useful description of a need for design in service process delivery.      See also Service Design.

" ... SCAD Service Design students proudly (or should I say, courageously...) present the first product based on the Service Design Taxonomy. 

Actually, it all started with an analysis that tried to map all functions that a service blueprint could facilitate. One of the most interesting ones was "facilitating cross-functional communication in support of customer-focused solutions" (Bitner, Ostrom 2008). That it is why it was named "huddle," the Service Design Huddle.

Basically, it is a workshop led by Service Designers and promotes a conversation with a group of stakeholders about an existing or new service proposition. This conversation should have from two to four hours duration and discuss each one of the 68 words that compose the taxonomy. The main goal is to select a "size" for each word, represented by a folding paper card, following a specific category sequence. The size of the card  (1/1, 1/2 or 1/4) should be selected based on the consideration or effort that the participants believe is or will be exerted to that word within the specific service. ... "