EN / DE / IT / SL / EL 

AIAE » E-LEARNING » INTERVIEW 7

Machine learning

The following interview snippets were given by Fabio Del Frate and cover the definition of AI, the most important aspects of machine learning, the interpretations about deep learning, the topics of supervised and unsupervised learning and its relation with AI as well as the different types of machine learning.

transcript

Fabio Del Frate received from the University of Rome “Tor Vergata” a Master Degree in Electronics Engineering (1992) and the Ph.D. degree in Computer Science (1997). In 1995-1996 he was Visiting Scientist at Massachusetts Institute of Technology, and from 1998 to 1999 he was with European Space Agency. He then joined the University of Rome “Tor Vergata”, where he is currently Associate Professor of Remote Sensing and Applied Electromagnetism in various Master and PhD Programs. On those topics, he has been lecturer at several European Universities, as well as principal investigator / project manager in several European Space Agency (ESA) and Italian Space Agency (ASI) funded projects, leading research activities in the field of Artificial Intelligence applied to satellite data for Earth Observation (referred to as “EO” in the following).

Quiz question 1/8

The interviewee, Fabio Del Frate, is a doctor of computer science, currently employed as a professor at the University of Rome Tor Vergata. He lectures about remote sensing and electromagnetics. In European and Italian Space Agencies he was leading research activity in the field of artificial intelligence applied to satellite data for Earth observations.


Quiz question 1/1

You have completed the Quiz!

Loading...

transcript

Q1. Prof. Del Frate, let’s starts from basics. What is Artificial Intelligence?

Well, to shortly explain what AI is I often like to cite one of the fathers of AI, Prof Nils John Nilsson. For him  “Artificial intelligence is that activity devoted to making machines intelligent, and intelligence is that quality that enables an entity to function appropriately and with foresight in its environment.” 

So, we can say that we provide to the machines and computers more degrees of freedom. We’re asking them to make their own decisions to go beyond what has been happening with traditional computer programming, that is to basically execute a sequence, even if long and complex, of command lines. 

How do we realize this ? It basically consists in feeding the machines with data and in providing them the capability to look for relationships, even if very subtle, among these data. It is with these two elements: data and  models for interpretation, that AI develops and builds its own knowledge, its own experience

I think that in this context an historical perspective could be also interesting, so that we might be curious to know when the first artificial intelligence and machine learning algorithms have been proposed. There might be different answers but my opinion is that to find the roots of artificial intelligence we should go back to 1943 when the first mathematical model of neural networks was presented in the scientific paper “A logical calculus of the ideas immanent in nervous activity” by Walter Pitts and Warren McCulloch. In this paper for the first time a single neuron is presented as an elemental nonlinear processing unit and it is assumed to be connected to other neurons. In other words a first rough but effective mathematical representation of brain.

Now the point is: is all this dangerous ? I think the answer is “Yes, it might be”, but just as much as it would be, or it has been, every time we’re fascinated by a new technology and we want to let it grow. Actually, the issue of developing a “safe” AI is definitely to the attention of the communities of scientists and researchers. The key point is that human beings should always act as supervisors. The rules, the guidelines, but most of all, the data on which the machine, as we said before, builds its knowledge, on which it will rely when taking its decisions, are provided by human experts, so that the machine yes, behaves with a certain level of autonomy, but within a well fixed playground, and the borders of this playground are decided by human beings. 

Quiz question 1/8

AI is providing machines and computers more degrees of freedom, meaning that machines are asked to go beyond the traditional computer programming. In order for machines to identify patterns, relationships, and build their own knowledge, it is needed to provide them data. Two important elements for interpretation of results in AI are data and models. AI has both positive and inspiring results, but it can also prove to be dangerous. Therefore, researchers must focus on safe development of artificial intelligence. The main principle must be that a human being is always in the role of a supervisor of rules and guidelines, but especially data.


Quiz question 1/1

You have completed the Quiz!

Loading...

transcript

Q2. What is Machine Learning?

Well, I would say that machine learning is a possible, I guess at the moment the most used one, technical process to generate artificial intelligence inside a machine. We’ve seen already that there are  two main actors: the data and the model, more exactly a mathematical model. The challenge is to find the parameters of the mathematical model so that it can represent the relationships among the data. Everything relies on setting-up a learning process based on the data themselves. We may have different types of models, for example neural networks or support vector machines and, consequently different algorithms to learn the model. The final goal however is always the same: to extract knowledge from the training data used in the training process and to apply this extracted knowledge to new data. By the way, the evaluation of how good the performance of the model is on the new data, that is data not considered during the training phase, is very important. It says to us how much the model is applicable in the real world, how much it is able to generalize. 

In any case, the quality and the quantity of the data used during the learning phase are crucial factors for a successful performance. In particular, the quantity of data needs to be statistically significant to provide the relevant features of the phenomenon we want to investigate. Another important key issue is to decide how complex our mathematical model should be. It would be a serious problem if we choose a too simple model to look for very subtle relationships among the data but, at the same way, if the rules underlying the data are less complicated, also the structure of the model can be lighter, sometime…. less is more, as we say, where less is referred to the  number of parameters characterizing the model. 

By the way, the model complexity  has also to do with the time we need to train the model and with the amount of data which is necessary for the implementation of the complete learning process.

Quiz question 1/8

What is machine learning?





Quiz question 1/8

What does machine learning do?





Quiz question 1/8

What are the results of different types of models?





Quiz question 1/8

Why is it important to evaluate the performance of the model on new data?





Quiz question 1/8

What are crucial factors for successful performance of AI?





Quiz question 1/1

You have completed the Quiz!

Loading...

transcript

Q3. AI & ML are often associated to terms like “deep learning”. What is Deep Learning?

Well, we may say that the word deep in deep learning has two different meanings, although they are very much correlated. One has to do with the hierarchical representation of the scenario we’re interested in.  If the scenario is for example object recognition, deep has to do with the multiple stages in the process of recognizing the object: as an animal, as a dog, as a Dalmatian dog.

These multiple stages of representation find a correspondence with the different layers of the mathematical model that we want to use to represent it, so, to simplify, a layer of the mathematical model focuses on the animal, a layer on the dog, a layer on the type of dog.  This is what happens with a deep neural network. Neural Networks are the most used models for deep learning. So here we have the second meaning, which is a topological meaning, addressing the number of layers of the architecture. So we have deep neural networks, such as Convolutional Neural Networks, which involve many processing layers against shallow neural networks, such as conventional Multi-Layer Perceptrons, which are neural networks with a limited number of layers. To go back to what we said about the complexity of the models, deep learning architectures are undoubtedly very complex, which means high computational burden required, and availability of huge quantities of learning data. Indeed, the need of a very high number of data stimulated the development of new techniques for what is called “data augmentation”, which consists in increasing the number of data starting from a more limited data set.

Quiz question 1/8

Yes
No
 

Can we say that deep learning has two meanings?


Is deep learning responsible for multiple stages in the process of recognising objects?


Does a mathematical model contain different layers of knowledge?


Is it true that neural networks are not the most used networks for deep learning?


Is one of the meanings of deep learning the topological one?


Neural networks can be shallow or convolutional?


Is deep learning a non-complex strategy?



Quiz question 1/1

You have completed the Quiz!

Loading...

transcript

Q4. What is Supervised Learning? Can you give some examples of applications that use it?

Well I would say that Supervised learning is the most common sub-branch of machine learning today. Supervised learning is a form of learning where, for a given input, a “true” outcome is available and this truth can be taught to the model. In other words, as a teacher or a supervisor, for each input given to the model we know what the output of the model should be and we want the model to incorporate the general rule. Typically, this happens by minimizing what is called an error or a cost function, which represents the distance between the desired output values and the actual output values generated by the model during the training phase. Actually, supervised learning can be assumed to be successful when the final overall error over a set of examples, which have not been used during the training phase, is under a predefined threshold. The training phase can be very long, but once we have concluded it properly, we have a very powerful tool, basically operating in real time, for the processing of the new data. Supervised learning can be useful in many fields. A rather massive application regards image classification. In this case the true data are data labelled by human image-interpretation. The classification can be performed at various levels: pixel level, patch level, object level. The level of classification guides the choice of the input that we want to give to the model. More in particular, supervised learning is now very much applied in Earth Observation where we want to make estimation of bio-geophysical parameters from data collected by satellites. An interesting example regards precision farming in agriculture. If we want to monitor the biomass of a crop in order to optimize the amount of fertilizer to be used, we can exploit satellite measurements taken on that crop. However, especially at the microwave region of the electromagnetic spectrum, the extraction of vegetation parameters is not trivial because the measurement is also, let’s say, contaminated by the underlying soil, so AI supervised models can definitely be one of the tools which allow us to retrieve that part of the information in the measurement more connected to the vegetation, discarding the one more conditioned by the soil.

Quiz question 1/8

True or false?

Supervised learning is not that common sub-branch of machine learning.




Quiz question 1/8

True or false?

Error represents the distance between the decided and actual output values.




Quiz question 1/8

True or false?

The training phase of supervised learning is short.




Quiz question 1/8

True or false?

The end result of supervised learning is very powerful, because it can operate on new data in real time.




Quiz question 1/8

True or false?

Image classification and object detection are examples of supervised learning.




Quiz question 1/8

True or false?

Supervised learning is rarely applied to earth observations.




Quiz question 1/8

True or false?

Biophysical parameters from data collected by satellites are included in supervised learning when earth observations are made.




Quiz question 1/1

You have completed the Quiz!

Loading...

transcript

Q5. What is Unsupervised Learning? Can you give some examples of applications using it?

The main purpose of unsupervised learning is to find correlations among the data so as to regroup them in similar categories. In unsupervised learning we don’t need external labelling operations or a teacher providing target values. The entire training of the mathematical model is based only on the existent original data so in this case, with respect to supervised learning, the process is characterized by a higher level of automation. We can say that, at the end of the training, the model organizes an internal representation of the data which make them more understandable and easy to be interpreted. As for the other learning paradigms, also unsupervised learning can be implemented by means of  various mathematical models. Also in this case, we need always to keep an eye to the complexity of the model that should reflect the complexity of the internal representation that we’re looking for. In EO and in other fields unsupervised learning can be used for dimensionality reduction. If we consider an hyperspectral measurement, we have that the collected information may be distributed over hundreds of wavelengths. In many cases it would be desirable to have such information, or the most significant of it, compressed in a fewer number of components. This can be achieved with unsupervised learning techniques.

Quiz question 1/8

1) What are we looking for in data with unsupervised learning? Correlation.
2) Based on what is regrouping of data done in unsupervised learning? Similarities.
3) Are external labelling operations needed in unsupervised learning? No.
4) What is the training of mathematical models based on? Original data.
5) What does the model do with the data at the end of the training in unsupervised learning? It organises it.
6) What do we need to pay attention to in unsupervised learning? Complexity of model.
7) For what in particular can unsupervised learning be used? Dimensionality reduction.


Quiz question 1/1

You have completed the Quiz!

Loading...

transcript

Q6. There are various groupings of Machine Learning algorithms, e.g. regression, classification, and so forth. Can you explain their relevance and difference between them?

There are various possibilities to group machine learning algorithms. I think that the main one relies on the type of the mathematical model. More in particular, as mentioned before, we have models for supervised training and models for unsupervised training. Regarding the former, among the most important ones we have neural networks, deep or shallow, support vector machines, decision trees, ensemble techniques. For the unsupervised models we can mention k-means, principal component analysis, auto-associative neural networks, self-organized maps, and so forth. 

But as your question was pointing out, another possibility to group machine learning algorithms addresses the distinction between regression and classification. Once a certain input is provided to our mathematical model, the difference between regression and classification mainly relies on the type of output we’re interested in. In the case of classification, the output will be a label, a class. For example, let’s consider again a satellite application where remote sensing is used for meteorology. Let’s therefore assume we have to make a meteorological analysis, and we want to decide which pixel in an image of the sky has to be associated to a cloudy sky and which pixel of the image to clear sky, so that the final purpose is to produce a map where we represent with the blue color the clear sky pixels and with the grey color the cloudy pixels. We say in this case that we want to obtain a classification map. For the regression, basically, we want that our mathematical model, as a response of its data processing, computes, or estimates, a number, more in general a real value number. So, if we consider the same scenario set on meteorology, we want to know from the remotely sensed data how much is the humidity or the precipitation rate, real numbers, in a given area. However, it is interesting to note how a regression problem can be translated into a classification problem. This is what happens if I say: within this range of precipitation values, so between these two numbers, I have the class “low precipitation”, in this other interval or range I have “moderate precipitation”, and so on. In this case I map ranges of values into classes.

Thank you, Prof. Fabio Del Frate, we really appreciated. It has been a clear & interesting snapshot of Artificial Intelligence features and relevant applications.

Quiz question 1/8

True or false?

There are only two different machine learning algorithms.




Quiz question 1/8

True or false?

Neural networks algorithms can be deep or shallow.




Quiz question 1/8

True or false?

Support-vector machines and decision trees are algorithms for supervised machine learning..




Quiz question 1/8

True or false?

Self-organising maps, means clustering and principal component analysis are types of supervised machine learning.




Quiz question 1/8

True or false?

The main difference between classification and regression is in the type of output.




Quiz question 1/8

True or false?

Output in classification is a label/class.




Quiz question 1/8

True or false?

Regression estimates a real number.




Quiz question 1/8

True or false?

We can’t transfer the regression problem into a classification one.




Quiz question 1/1

You have completed the Quiz!

Loading...

The learning unit has not been completed

You have successfully completed the learning unit.