Elearning! December 2015 / January 2016 9
News
Deep Learning Technology
Moves Onward and Upward
The concept of "deep learning" took a giant step forward last month
when (1) Google released to the open-source community the machine
learning system at the core of many of its services, and (2) Rice Analytics
released its automated RELR deep machine-learning tool to the masses
as a Cloud Web application and an API.
Deep learning, or machine learning, involves using computer proces-
sors that act similar to neurons in the human brain.
TENSORFLOW'S IMPACT
Google's artifcial intelligence engine TensorFlow — is said to be fve
times faster than any of its previous versions. Company CEO Sundar
Pichai says that it shows promise for when researchers "are trying to
make sense of very complex data — everything from protein folding
to crunching astronomy data." While performing these operations, the
"neural networks" discover patterns and relationships as they absorb all
the available information.
By open-sourcing TensorFlow, Google is able to tap the broader devel-
oper and scientifc community to vastly improve the source code, which
in turn might help the company improve its own products. The chips were
designed initially for rendering the graphics of the games, but engineers
eventually found that it had many other uses as well. Many of Google's
tools have already use a few versions of "artifcial intelligence (A.l.) to
work, such as voice-to-text, language translation and Web search.
From using TensorFlow A.I. to flter out spam in g-mail to assisting
faster discovery of life-saving drugs, Google has been applying the deep
learning technology in a number of ways.
PROVIDING EXPLANATIONS
St. Louis-based Rice Analytics says its automated RELR deep machine-
learning tool differs from conventional A.I. approaches by providing ex-
planations for the reasons behind its predictions. A Cloud-based version
of the tool called SkyRELR is said to allow machine learning that extends
beyond other forms of A.I. and neural networks by drawing on "deeper
patterns that are hidden to other forms of artifcial intelligence."
The other selling point, the startup says, is transparency: "Deep learn-
ing is not a black box," the company stresses.
RELR (pronounced "reller") is short for "reduced error logistics re-
gression," which is described as a neuromorphic algorithm designed to
model the deep explicit and implicit learning mechanisms of neurons.
Rice Analytics says its secret is leveraging an ability to provide "explana-
tory models" in the form of explicit models that can be used to interpret
the reasons for predictions.
SkyRELR is touted as yielding models that are equivalent to a four-
layer deep-learning artifcial neural network.
Now that the new deep-learning technology has been wrung out in
targeted advertising campaigns, the Cloud-based version may open up
the technology to other enterprise customers who are not data scientists.
—More info: http://or-politics.com/uncategorized/google-open-sourc-
es-its-tensorfow-deep-learning-engine/150519/; http://thenewstele-
graph.com/2015/11/15/googles-tensorfow-machine-learning-system-
is-now-open/; www.datanami.com/2015/11/17/cloud-based-deep-
learning-tool-explains-predictions/
Talent Pros' Salaries Unchanged
In 2015, the median annual salary for someone
in the talent development profession is between
$70,000 and $79,999. The median salary
is unchanged from 2011.
Key fndings in a new report from the ATD:
>> Most talent development professionals
have at least a four-year degree and
nearly half (49 percent) have a master's
or advanced degree.
>> Median salary in the U.S., between $70,000 and $79,999, remains
unchanged since 2011.
>> 56 percent have a base salary between $50,000 and $89,999.
>> Pay tends to be higher in the Northeast and West.
>> Participants who listed training delivery as their primary respon-
sibility made less that those who named other primary areas of
responsibility (instructional design or change management, for
example).
>> 75 percent of respondents received a raise the previous year, typi-
cally less than 4 percent of base salary.
>> 56 percent receive incentive rewards, but these make up less than
10 percent of compensation.
>> 88 percent were offered medical coverage by their employer.
—More info: www.td.org/salaryreport
Do You Know a Learning Champion?
Elearning! Media Group's new Learning! Champion Award will
recognize individuals for exceptional contributions to the industry.
Nominees can be innovators, thought-leaders, trail-blazers, mentors,
cutting-edge technologists, creative consultants, or high-performing
learning leaders that positively impact the learning industry.
Learning! Champion award-winners will be featured in Elearning!
magazine's May 2016 issue and invited to join the Learning! Cham-
pion Award Reception at the Enterprise Learning! Conference 2016.
A committee of practitioners, analysts and magazine editors will take
part in evaluating nominations. A validation process will be conducted
as well as relevance testing. After which, nominees will be selected
and ranked within each category.
Nominating a candidate is easy via the online ballot and takes only
minutes to apply. There is no charge. A candidate can be submit-
ted only one time per year for consideration. Upon completion of the
application, you will receive via email a confrmation of receipt along
with a copy of the nomination. All applications must be received by
Dec. 31.
While self-nominations are not explicitly prevented by the rules,
they are somewhat discouraged.
Email questions to awards@2elearning.com for review.
—More info: www.2elearning.com/awards/learning-champion-award