Research Associates

Report info

Report date	Tuesday, May 17, 2011 2:57:04 PM BST

Start date	Monday, April 4, 2011 12:47:00 AM BST

Stop date	Sunday, April 24, 2011 11:59:00 PM BST

Stored responses	127

Number of completed responses	107

Question 1

Do you conduct research in, teach, or study machine learning?

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Yes	122	96.06%	96.06%
No	5	3.94%	3.94%
Sum:	127	100%	100%
Total answered: 127

Question 2

What do you do?

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Research associate or postdoctoral fellow at an academic institution	127	100%	100%
Sum:	127	100%	100%
Total answered: 127

Question 3

Which department(s) are you based in? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Biology	7	3.7%	5.51%
Cognitive science	7	3.7%	5.51%
Computer science	84	44.44%	66.14%
Electrical engineering	10	5.29%	7.87%
Engineering sciences	11	5.82%	8.66%
Machine Learning	30	15.87%	23.62%
Mathematics	6	3.17%	4.72%
Neuroscience	15	7.94%	11.81%
Operations research	1	0.53%	0.79%
Physics	2	1.06%	1.57%
Psychology	1	0.53%	0.79%
Statistics	10	5.29%	7.87%
Other:	5	2.65%	3.94%
Sum:	189	100%	100%
Total answered: 127

Text input

bioinformatics

Linguistics

Biophysics

Veterinary bioscience

Signal and Communications Theory

Question 4

Which conferences do you regularly attend? (You may pick more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
AAAI	15	5.08%	15%
ACL	13	4.41%	13%
AISTATS	22	7.46%	22%
ALT	3	1.02%	3%
CogSci	2	0.68%	2%
COLT	8	2.71%	8%
CoNLL	8	2.71%	8%
Cosyne	5	1.69%	5%
CVPR	8	2.71%	8%
ECCV	3	1.02%	3%
ECML	26	8.81%	26%
EMNLP	12	4.07%	12%
ICCV	6	2.03%	6%
ICML	47	15.93%	47%
IJCAI	16	5.42%	16%
IROS	2	0.68%	2%
ISMB	3	1.02%	3%
NAACL	5	1.69%	5%
NIPS	65	22.03%	65%
SIGIR	2	0.68%	2%
Snowbird	4	1.36%	4%
UAI	20	6.78%	20%
Sum:	295	100%	100%
Total answered: 100

Text input

Society for Neuroscience meetings

KDD

ISMIR

ICASSP, Interspeech

OHBM

WWW

MGED
NetSCI

Benelearn, Benelux Meeting on Systems and Control

WI
ECML

KDD

ISBA

Interspeech, HLT-ACL, SigDial

MLSP

ComputationWorld, Text Analysis Conference (TAC)

KDD

R:SS, ISRR and ICRA are better conferences than IROS which I attend more frequently.

ICGI

ICC

SAB,ICDL,EPIROB

MSML

INTERSPEECH, ICASSP

IEEE CDC, MTNS

CBMS, ASMDA

ICANN

SIGKDD

ICASSP, MLSP

GECCO
CEC

ICMLA, IDEAL

Question 5

Do you have (or have you had) students in machine learning?

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Yes	49	38.58%	38.58%
No	78	61.42%	61.42%
Sum:	127	100%	100%
Total answered: 127

Question 6

What is your educational background/training/disciplinary background? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Cognitive science	10	4%	8%
Computer science	99	39.6%	79.2%
Economics	2	0.8%	1.6%
Electrical engineering	24	9.6%	19.2%
Engineering sciences	10	4%	8%
Machine Learning	40	16%	32%
Mathematics	26	10.4%	20.8%
Neuroscience	8	3.2%	6.4%
Operations research	1	0.4%	0.8%
Physics	9	3.6%	7.2%
Psychology	3	1.2%	2.4%
Social science	2	0.8%	1.6%
Statistics	10	4%	8%
Other:	6	2.4%	4.8%
Sum:	250	100%	100%
Total answered: 125

Text input

Linguistics

Logic-based artificial Intelligence

Practical philosophy / economics / political science

Mechanical Engineering

Philosophy

electronics

Question 7

Did your educational background prepare you well for research in machine learning?

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Poorly prepared)	11	8.66%	8.8%
2	21	16.54%	16.8%
3	63	49.61%	50.4%
4 (Well prepared)	30	23.62%	24%
Not answered:	2	1.57%	-
Sum:	127	100%	100%
Total answered: 125

Question 8

What educational backgrounds do your students have? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Biology	5	4.46%	10.64%
Cognitive science	2	1.79%	4.26%
Computer science	40	35.71%	85.11%
Economics	3	2.68%	6.38%
Electrical engineering	14	12.5%	29.79%
Engineering sciences	10	8.93%	21.28%
Machine Learning	7	6.25%	14.89%
Mathematics	11	9.82%	23.4%
Neuroscience	2	1.79%	4.26%
Physics	6	5.36%	12.77%
Psychology	3	2.68%	6.38%
Statistics	8	7.14%	17.02%
Other:	1	0.89%	2.13%
Sum:	112	100%	100%
Total answered: 47

Text input

Biomedical engineering

Question 9

Did the educational backgrounds of your students prepare them well for studying machine learning?

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Poorly prepared)	2	1.57%	4.26%
2	23	18.11%	48.94%
3	20	15.75%	42.55%
4 (Well prepared)	2	1.57%	4.26%
Not answered:	80	62.99%	-
Sum:	127	100%	100%
Total answered: 47

Question 10

In your opinion, what will constitute good educational backgrounds for a graduate student entering machine learning? (You may select as many as you like).

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Biology	2	0.37%	1.67%
Cognitive science	18	3.31%	15%
Computer science	107	19.71%	89.17%
Economics	5	0.92%	4.17%
Electrical engineering	19	3.5%	15.83%
Engineering sciences	20	3.68%	16.67%
Machine Learning	100	18.42%	83.33%
Mathematics	106	19.52%	88.33%
Neuroscience	10	1.84%	8.33%
Operations research	29	5.34%	24.17%
Physics	31	5.71%	25.83%
Psychology	2	0.37%	1.67%
Social science	1	0.18%	0.83%
Statistics	93	17.13%	77.5%
Sum:	543	100%	100%
Total answered: 120

Text input

I think that some of the very relevant lecture courses are offered in a variety of disciplines-- so, it might be more a question of "what lectures did they attend" rather than "what degree did they get in the end".

Really any of these could work, it seems to me that the important thing is to have some kind of mathematical and computational background, and to have been exposed to the issues involved in working on problems with a large number of variables.

The most essential things are mathematics and programming skills.

All the fields might be highly relevant depending on what you want to do with machine learning. I marked the ones which I think are the most important ones.

Good teaching is important.

Most computer science and engineering programs do not seem to produce good machine learning researcher candidates. Instead, the best ones typically come with math/physics/statistics background.

I believe that one can enter machine learning from many directions. What is important is some mathematical maturity, and the willingness to get one's hands dirty in several things, e.g. proving some theory, writing some code, and modeling some real world problem. Backgrounds that prepare students for either one of the three areas above would work well.

Learning methods
computer-human learning

Good mathematics and computer science background really help students in entering machine learning

I think Biology, Social science and Neuroscience are really application areas
Economics and Psychology have some value in understanding subjective probability, utility etc...

The other areas are highly overlapping with each other and all provide a good background in machine learning....

Ranking: Statistics, Mathematics, Electrical Engineering
(I'm not quite sure what the curriculum in machine learning would all contain. But obviously, it should be a good preparation!)

Maths, Maths and Maths. And statistics. I don't have a solid background in it and thus need to catch-up everything now. If I had pure mathematics foundations, this would greatly benefit my understanding and development of machine learning methods and applications.

Question 11

What area(s) in machine learning do you work in? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Bayesian statistics	46	8.98%	41.82%
Control and planning	16	3.12%	14.55%
Deep learning	5	0.98%	4.55%
Dimensionality reduction	27	5.27%	24.55%
Graphical models	43	8.4%	39.09%
Information theory	11	2.15%	10%
Kernel methods	24	4.69%	21.82%
Learning theory	18	3.52%	16.36%
Manifold learning	4	0.78%	3.64%
Model selection	18	3.52%	16.36%
Neural networks	23	4.49%	20.91%
Nonparametric methods	20	3.91%	18.18%
Online learning	27	5.27%	24.55%
Optimization	26	5.08%	23.64%
Reinforcement learning	27	5.27%	24.55%
Relational learning	13	2.54%	11.82%
Semisupervised learning	21	4.1%	19.09%
Sparse learning	19	3.71%	17.27%
Statistical physics of learning	2	0.39%	1.82%
Structured learning	16	3.12%	14.55%
Supervised learning	43	8.4%	39.09%
Time series modelling	20	3.91%	18.18%
Unsupervised learning	40	7.81%	36.36%
Other:	3	0.59%	2.73%
Sum:	512	100%	100%
Total answered: 110

Text input

Multitask learning

Preference Learning

multistrategy ML

Question 12

What other area(s) do you work in? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
None	5	1.27%	4.55%
Artificial intelligence	31	7.89%	28.18%
Bioinformatics	22	5.6%	20%
Brain, medical imaging	13	3.31%	11.82%
Cognitive science	12	3.05%	10.91%
Collaborative filtering	4	1.02%	3.64%
Communications	4	1.02%	3.64%
Computer vision	12	3.05%	10.91%
Computer science	31	7.89%	28.18%
Databases	4	1.02%	3.64%
Data mining	23	5.85%	20.91%
Electrical engineering	5	1.27%	4.55%
Genetics	4	1.02%	3.64%
Multimedia processing	4	1.02%	3.64%
Linguistics	6	1.53%	5.45%
Image processing	7	1.78%	6.36%
Information retrieval	24	6.11%	21.82%
Mathematics	6	1.53%	5.45%
Natural language processing	20	5.09%	18.18%
Network analysis	8	2.04%	7.27%
Neural data analysis	9	2.29%	8.18%
Neuroprosthetics	3	0.76%	2.73%
Neuroscience	21	5.34%	19.09%
Operations research	5	1.27%	4.55%
Physics	1	0.25%	0.91%
Psychology	6	1.53%	5.45%
Robotics	13	3.31%	11.82%
Signal processing	12	3.05%	10.91%
Social science	10	2.54%	9.09%
Speech processing	6	1.53%	5.45%
Statistics	23	5.85%	20.91%
Systems biology	7	1.78%	6.36%
Text and web analysis	26	6.62%	23.64%
Other:	6	1.53%	5.45%
Sum:	393	100%	100%
Total answered: 110

Text input

music

Computational Linguistics

knowledge innovation

stochastic mechanics

botany

Biomedical science

Question 13

In your opinion, what are core (foundation) areas of machine learning? (You may select as many as you see fit)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Bayesian statistics	71	8.79%	64.55%
Control and planning	18	2.23%	16.36%
Deep learning	13	1.61%	11.82%
Dimensionality reduction	31	3.84%	28.18%
Graphical models	52	6.44%	47.27%
Information theory	48	5.94%	43.64%
Kernel methods	42	5.2%	38.18%
Learning theory	71	8.79%	64.55%
Manifold learning	10	1.24%	9.09%
Model selection	37	4.58%	33.64%
Neural networks	26	3.22%	23.64%
Nonparametric methods	29	3.59%	26.36%
Online learning	33	4.08%	30%
Optimization	56	6.93%	50.91%
Reinforcement learning	42	5.2%	38.18%
Relational learning	7	0.87%	6.36%
Semisupervised learning	23	2.85%	20.91%
Sparse learning	17	2.1%	15.45%
Statistical physics of learning	11	1.36%	10%
Structured learning	21	2.6%	19.09%
Supervised learning	65	8.04%	59.09%
Time series modelling	18	2.23%	16.36%
Unsupervised learning	58	7.18%	52.73%
Other:	9	1.11%	8.18%
Sum:	808	100%	100%
Total answered: 110

Text input

all of them

Artificial Intelligence

Preference learning

-

High dimensional integration

No clear opinion, sorry

pattern recognition

multi dimensional statistics

Real Analysis

Question 14

In your opinion, machine learning is a subfield of statistics.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	16	12.6%	14.81%
2	26	20.47%	24.07%
3	42	33.07%	38.89%
4 (Agree)	24	18.9%	22.22%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 15

In your opinion, machine learning is a subfield of computer science.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	10	7.87%	9.26%
2	18	14.17%	16.67%
3	48	37.8%	44.44%
4 (Agree)	32	25.2%	29.63%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 16

In your opinion, machine learning is a subfield of cognitive science.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	52	40.94%	48.15%
2	35	27.56%	32.41%
3	17	13.39%	15.74%
4 (Agree)	4	3.15%	3.7%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 17

In your opinion, machine learning is a subfield of neuroscience.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	60	47.24%	55.56%
2	37	29.13%	34.26%
3	10	7.87%	9.26%
4 (Agree)	1	0.79%	0.93%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 18

In your opinion, machine learning is a subfield of engineering.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	25	19.69%	23.15%
2	42	33.07%	38.89%
3	29	22.83%	26.85%
4 (Agree)	12	9.45%	11.11%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 19

In your opinion, machine learning is a subfield of artificial intelligence.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	8	6.3%	7.41%
2	15	11.81%	13.89%
3	46	36.22%	42.59%
4 (Agree)	39	30.71%	36.11%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 20

In your opinion, machine learning is an interdisciplinary area.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	2	1.57%	1.85%
2	7	5.51%	6.48%
3	29	22.83%	26.85%
4 (Agree)	70	55.12%	64.81%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 21

In your opinion, machine learning is a a discipline distinct from other disciplines.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	18	14.17%	16.67%
2	40	31.5%	37.04%
3	26	20.47%	24.07%
4 (Agree)	24	18.9%	22.22%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 22

Any further comments regarding the relationship of machine learning to other disciplines?

Text input

I think that the answer to this question depends a lot on whether on defines it by "content", i.e. by how the topics researched in machine learning overlap with those in other fields, or by "community", i.e. how machine learning overlaps with other disciplines either by things like peoples' background and publication habits. If defined purely by research goals, an argument could be made that machine learning is really computational statistics with a bigger focus on applications, and with some artificial intelligence thrown in as a long-term goal... Also, there is a small overlap with some areas of econometrics which is a bit underappreciated in my opinion.

Machine learning is a descendant of statistics, computer science, and engineering. It deals with algorithms that try to make accurate predictions from data.

I found that you mainly ignored the connection between ML and data mining (KDD, WWW, etc) giving that the communities are converging to the same methods and applications.

As Poggio once said, as much as deciphering the genetic code was the big challenge in the
last years, understanding intelligence is the next challenge. Machine learning is the key to do that.

Machine learning needs (a) a solid mathematical background (b) solid knowledge of computer algorithm engineering (c) feel comfortable with handling/programming computers and (d) be able to understand and focus to particular field areas, such as visual/auditive/linguistic perception, bioinformatics,...

I don't consider machine learning to be a "subfield" of statistics, because its very broad in scope, and it didn't really come out of statistics historically. I do think that in terms of the work being done that statistics overlaps with machine learning much more than any other field though.

So, I understood the expression "subfield" similar to "subset", meaning, a subfield does not have elements outside one of the above mentioned disciplines. This is why I disagreed with most of those statements.

"Machine learning is a subfield of AI" could also be given as the opposite statement "AI is a subfield of ML".

I tend to see ML as a "geeky" side of stats/AI: strong emphasis on implementations, on benchmarking of algorithms, on "it works, anyways!" combination of several algorithms -- as opposed to more abstract considerations on mathematical foundations. Maybe is my opinion biased by the conferences I attend and review for.

There are many facets to machine learning so simple subset relationships are not possible.

human learning, and machine learning with interaction with human (computer learns from human)
Goal oriented learning - computer need to learn to help human

I believe Machine Learning is an area between Mathematics and Computer Science.

machine learning overlaps with all the mentioned areas in previous question but is not strictly a subfield of them. More of a union of a lot of subfields from different disciplines.

I personally felt that machine learning is just a tool for solving various real problems. A machine learning algorithm itself would not be meaningful if there is no applications to it.

Question 23

In your opinion, in the university, machine learning should ideally be a group within computer science.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	5	3.94%	4.63%
2	20	15.75%	18.52%
3	55	43.31%	50.93%
4 (Agree)	28	22.05%	25.93%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 24

In your opinion, in the university, machine learning should ideally be a group within neuroscience.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	60	47.24%	55.56%
2	35	27.56%	32.41%
3	13	10.24%	12.04%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 25

In your opinion, in the university, machine learning should ideally be a group within engineering.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	34	26.77%	31.48%
2	42	33.07%	38.89%
3	23	18.11%	21.3%
4 (Agree)	9	7.09%	8.33%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 26

In your opinion, in the university, machine learning should ideally be a group within statistics.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	20	15.75%	18.52%
2	37	29.13%	34.26%
3	37	29.13%	34.26%
4 (Agree)	14	11.02%	12.96%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 27

In your opinion, in the university, machine learning should ideally be a group within psychology.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	81	63.78%	75%
2	25	19.69%	23.15%
3	2	1.57%	1.85%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 28

In your opinion, in the university, machine learning should ideally be a spread across multiple departments.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	17	13.39%	15.74%
2	28	22.05%	25.93%
3	39	30.71%	36.11%
4 (Agree)	24	18.9%	22.22%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 29

In your opinion, in the university, machine learning should ideally be an interdisciplinary centre.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	4	3.15%	3.7%
2	19	14.96%	17.59%
3	39	30.71%	36.11%
4 (Agree)	46	36.22%	42.59%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 30

In your opinion, in the university, machine learning should ideally be its own department.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	26	20.47%	24.07%
2	28	22.05%	25.93%
3	31	24.41%	28.7%
4 (Agree)	23	18.11%	21.3%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 31

In your opinion, in the university, machine learning should be taught at the graduate level.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	2	1.57%	1.85%
2	10	7.87%	9.26%
3	33	25.98%	30.56%
4 (Agree)	63	49.61%	58.33%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 32

In your opinion, in the university, machine learning should be taught as a specialist track in a related undergraduate programme.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	2	1.57%	1.85%
2	18	14.17%	16.67%
3	48	37.8%	44.44%
4 (Agree)	40	31.5%	37.04%
Not answered:	19	14.96%	-
Sum:	127	100%	100%
Total answered: 108

Question 33

In your opinion, in the university, machine learning should be taught in a dedicated undergraduate programme.

Frequency table
Levels	Absolute frequency	Relative frequency	Adjusted relative frequency
1 (Disagree)	18	14.17%	16.82%
2	51	40.16%	47.66%
3	25	19.69%	23.36%
4 (Agree)	13	10.24%	12.15%
Not answered:	20	15.75%	-
Sum:	127	100%	100%
Total answered: 107

Question 34

In your opinion, in which departments should machine learning be taught?

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
Biology	19	3.47%	17.59%
Cognitive science	32	5.84%	29.63%
Computer science	104	18.98%	96.3%
Economics	17	3.1%	15.74%
Electrical engineering	32	5.84%	29.63%
Engineering sciences	45	8.21%	41.67%
Machine Learning	81	14.78%	75%
Mathematics	39	7.12%	36.11%
Neuroscience	33	6.02%	30.56%
Operations research	30	5.47%	27.78%
Physics	11	2.01%	10.19%
Psychology	13	2.37%	12.04%
Social science	11	2.01%	10.19%
Statistics	78	14.23%	72.22%
Other:	3	0.55%	2.78%
Sum:	548	100%	100%
Total answered: 108

Text input

all

basically all of them

bioinformatics

Question 35

Any further comments regarding the place of machine learning within the university?

Text input

In my opinion, machine learning is a subfield of CS like any other -- like AI, Theory, etc., and should be treated as such.

Ideally, introductory courses/seminars could also be provided in all the unchecked, aforementioned departments. This is to indicate to other departments the possibility of using Machine Learning as a tool.

These questions are confusing because there are many ways to implement the ideas.

I think the current diversity where machine learning is positioned in different ways to different departments in different universities is ok. I think the interests of the group leaders affect the ideal position greatly...

I am not sure if students should focus on machine learning at the undergraduate level... I think it is a useful area for students of many backgrounds to have some knowledge of.

A machine learning course focused on exploratory data analysis, particularly graphical would have very general appeal and could probably replace a statistics course and would be of value to a student in almost any dicipline. On the other hand courses in Bayesian theory are most suited to students in quantitive diciplines as well as areas such as pschology and economics.

I believe Machine Learning should be taught in collaboration between computer science and mathematics, with access to the course/modules by other departments.

Question 36

What brought you into machine learning? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
I want to use machine learning techniques to analyse data.	67	28.03%	63.21%
I want to understand how the brain works.	28	11.72%	26.42%
I want to understand the fundamental principles of learning and intelligence.	55	23.01%	51.89%
I want to understand existing machine learning techniques and develop new ones.	50	20.92%	47.17%
I want to build intelligent computers or robots.	39	16.32%	36.79%
Sum:	239	100%	100%
Total answered: 106

Text input

In addition, I want to get a good job and make money and ML is a hot pre-requisite :)

I thought machine learning was well suited to biological data analysis, to extract meaningful signals from voluminous noisy data.

It is the dual case of logic-based AI

Randomness ;)

I like math.

I want to use machine learning for scientific discovery. Modern high throughput methods make manual discovery of new knowledge very hard.

From what I have seen, unfortunately machine learning is now far remote from its neuroscience origins (see e.g. NIPS).

Powerful statistic tool

To me it is fascinating how complex the brain is, yet we can develop simple and elegant mathematical equations that help us outperform the human brain in some, trivial yet surprisingly powerful, tasks.

everything is information

Question 37

In your opinion, what are the goals of machine learning? (You may select more than one)

Frequency table
Choices	Absolute frequency	Relative frequency	Adjusted relative frequency
To understand the fundamental principles of learning and intelligence.	66	19.47%	62.26%
To analyze and extract useful information from data.	88	25.96%	83.02%
To understand how the brain works.	24	7.08%	22.64%
To build smarter computers or robots.	64	18.88%	60.38%
To develop algorithms for making accurate predictions based on data.	97	28.61%	91.51%
Sum:	339	100%	100%
Total answered: 106

Text input

While some people argue that learning the fundamentals of learning is the goal (and maybe it should be, given the name of the field), most of the community seems to be working on the latter two goals

I think the other options are goals that people in machine learning can have, but generally when they also have a cross interest in another area as well - like neuroscience.

- To mathematically and computationally model real-world processes and data.

The last two (fundamental principles of learning and intelligence and how the brain works) needs to be done by (or in very, very close connexion) with neuroscientists, biologists, neurosurgeons, rather than algorithmic prone geeks ;) Otherwise, risk of coming up with a gazillion "ground-breaking" but not biology-backed (or on half-understood biology) schemes of "how the brain works". We (geeks) are experts at algorithms, coding, math, but not on biology -- and machine learning has evolved in our playground, getting further than its Neuroscience roots.

to help human in the tasks it performs better than human

Question 38

Do you have further comments regarding any part of this survey?

Text input

Now you can go ahead and apply learning algorithms over machine learners.

This is a very interesting effort.

Thanks for organizing it, looking forward to the picture of the field that will emerge from its results.

it is not necessary to put ML into the boxes
It is a multidisciplinary topic and shoud be developped and teached in many domains with different (adapted) points of view

I was surprised to see that when linking the fields that might relate to machine learning, statistics was included, yet sometimes mathematics was not.

Thanks for the great survey.