Reference Information
Title: A Few Useful Things to Know About Machine Learning
Author: Pedro Domingos
Citation: "A Few Useful Things to Know About Machine Learning", Pedro Domingos, Communications of the ACM, pp. 78-87, 2012.
Summary
This paper provided an overview of machine learning techniques at a broad scope. Machine learning consists of using learning algorithms with large amounts of training data to generalize the set of possible data in order to classify new, unknown data that is discovered. The focus of this paper was on classification algorithms, along with some key lessons in the field of machine learning regarding classification. Evaluating classifiers involves having a formal representation, an objective function for evaluation, and optimization. It should be noted that testing and training data should always be kept separate for more accurate evaluation statistics. This can be remedied by a few different solutions, including using cross validation.
Machine learning is all about generalizing the data that is given. A learner uses not only training data, but extra assumptions or knowledge about the domain, as well ("no free lunch" theorems). Some problems with the generalization include overfitting, underfitting, multiple testing, and the curse of dimensionality (generalizing becomes more difficult as more features are added into the input). Some theoretical guarantees that are incorrect were mentioned, including the fact that there are no exact guarantees on the bound of the number of examples needed, and that infinite data does not necessarily lead to a correct classifier. The importance of choosing the correct features was reiterated. While relative, independent features are the most useful, it is difficult to know what these features are when simply presented with the raw data of an input, since the input data tends to be more observational than experimental. In addition, it is more important to have large amounts of data for training purposes instead of having a more complex learning algorithm, but scalability must be taken into account when using lots of data. As for choosing the "best" learning algorithm, it really depends on the particular domain and application for which it is being used.
Finally, combining learning algorithms was discussed by using model ensembles to create more accurate learners from running the data through multiple classifiers. Some combination techniques include boosting, bagging, and stacking.
Thoughts
I have previously taken a course in machine learning, so I knew most of the information that was presented within this article. However, it was a good refresher for the information that I did know, and it provided some very good information regarding lessons and myths in machine learning that I did not previously know about.
For example, it was very interesting to learn about some of the details of possible machine learning problems, such as overfitting, multiple testing, and the curse of dimensionality. It was very useful that possible (and best) solutions were presented for each of these problems. Overall, the article was written in a very understandable format that made for an enjoyable and informative read.
Tuesday, March 26, 2013
Reading Assignment: The Word-Gesture Keyboard: Reimagining Keyboard Interaction
Reference Information
Title: The Word-Gesture Keyboard: Reimagining Keyboard Interactions
Authors: Shumin Zhai and Per Ola Kristensson
Citation: "The Word-Gesture Keyboard: Reimagining Keyboard Interactions", Shumin Zhai and Per Ola Kristensson, Communications of the ACM, pp. 91-101, 2012.
Summary
This paper discussed word-gesture, an alternative method of interaction for text input when using touch-screen keyboards, such as on mobile devices. The interaction consists of the action of swiping a single finger across the soft keyboard on a touch screen, running the finger consecutively across each letter of a word in one, fluid motion. It is designed to be a faster method of text input than using a traditional physical keyboard, but it still uses the same keyboard design so it is intended to be easy to learn and with the ability to improve usage speed with time. The faster speed of the word-gesture system stems from the fact that only a single continuous motion made with one finger is necessary to create a word and spaces are automatically inserted between the words. Ease of use comes from the fact that users are already familiar with the keyboard design, gestures come more naturally than the traditional usage of physical keyboards, and that no gestures are required to be learned since the user simply follows the pattern of keys that are visible on the screen.
The shape of the gesture that is created with this motion is compared against a set of pre-known gestures that are already associated with words in order to perform gesture recognition. The ability to enter commands (such as copy and paste) was added along with the ability to type words. Indexing and pruning are used to make the searching of the known gestures feasible for mobile devices.
The word-gesture method was tested through some experimentation along with releasing it as an application for mobile devices in order to receive feedback from real users of the system. The experiments included testing users on the ability to memorize gesture shapes and the speed of users' initial uses. Reviews made by users of the released application for mobile devices were used for evaluating the general conception of the system itself. One of the major contributions of word-gesture is that by releasing it as an actual product for people to use, the idea became more widespread, allowing for the proliferation of this new technique of text input.
Thoughts
It was really great to be able to read this paper about a recent system that is now in fairly widespread use in daily life. I have seen variations of this system on my own phone and therefore can relate my own experience with it to the information gained from reading this paper. Many of the papers that we read discuss systems that are not well known; however, the contribution of this paper is known by many now. This made it very interesting to learn how this method that I was previously aware of actually works.
I think that it is a very interesting concept to introduce a new method like this that relies mostly on previously-known concepts such as the physical keyboard. The fact that the user is not required to learn any new gestures, but simply to apply a new type of motion to a well-known system, is a very intriguing idea. It makes one start to think about what other types of new interaction can be applied to existing systems in order to improve upon their usage.
The focus on human psychology that was used in the creation and evaluation of the method was great to read about, since it was discussed why this method actually works. In addition, it was great to see that the main evaluation of the system occurred by putting it into real-world usage and obtaining actual reviews of the product, instead of simply running lab experiments to try and approximate real-world usage.
Title: The Word-Gesture Keyboard: Reimagining Keyboard Interactions
Authors: Shumin Zhai and Per Ola Kristensson
Citation: "The Word-Gesture Keyboard: Reimagining Keyboard Interactions", Shumin Zhai and Per Ola Kristensson, Communications of the ACM, pp. 91-101, 2012.
Summary
This paper discussed word-gesture, an alternative method of interaction for text input when using touch-screen keyboards, such as on mobile devices. The interaction consists of the action of swiping a single finger across the soft keyboard on a touch screen, running the finger consecutively across each letter of a word in one, fluid motion. It is designed to be a faster method of text input than using a traditional physical keyboard, but it still uses the same keyboard design so it is intended to be easy to learn and with the ability to improve usage speed with time. The faster speed of the word-gesture system stems from the fact that only a single continuous motion made with one finger is necessary to create a word and spaces are automatically inserted between the words. Ease of use comes from the fact that users are already familiar with the keyboard design, gestures come more naturally than the traditional usage of physical keyboards, and that no gestures are required to be learned since the user simply follows the pattern of keys that are visible on the screen.
The shape of the gesture that is created with this motion is compared against a set of pre-known gestures that are already associated with words in order to perform gesture recognition. The ability to enter commands (such as copy and paste) was added along with the ability to type words. Indexing and pruning are used to make the searching of the known gestures feasible for mobile devices.
The word-gesture method was tested through some experimentation along with releasing it as an application for mobile devices in order to receive feedback from real users of the system. The experiments included testing users on the ability to memorize gesture shapes and the speed of users' initial uses. Reviews made by users of the released application for mobile devices were used for evaluating the general conception of the system itself. One of the major contributions of word-gesture is that by releasing it as an actual product for people to use, the idea became more widespread, allowing for the proliferation of this new technique of text input.
Thoughts
It was really great to be able to read this paper about a recent system that is now in fairly widespread use in daily life. I have seen variations of this system on my own phone and therefore can relate my own experience with it to the information gained from reading this paper. Many of the papers that we read discuss systems that are not well known; however, the contribution of this paper is known by many now. This made it very interesting to learn how this method that I was previously aware of actually works.
I think that it is a very interesting concept to introduce a new method like this that relies mostly on previously-known concepts such as the physical keyboard. The fact that the user is not required to learn any new gestures, but simply to apply a new type of motion to a well-known system, is a very intriguing idea. It makes one start to think about what other types of new interaction can be applied to existing systems in order to improve upon their usage.
The focus on human psychology that was used in the creation and evaluation of the method was great to read about, since it was discussed why this method actually works. In addition, it was great to see that the main evaluation of the system occurred by putting it into real-world usage and obtaining actual reviews of the product, instead of simply running lab experiments to try and approximate real-world usage.
Monday, March 4, 2013
Reading Assignment: Sketch Based Interfaces: Early Processing for Sketch Understanding
Reference Information
Title: Sketch Based Interfaces: Early Processing for Sketch Understanding
Authors: Tevfik Metin Sezgin, Thomas Stahovich, Randall Davis
Citation: "Sketch Based Interfaces: Early Processing for Sketch Understanding", Tevfik Metin Sezgin, Thomas Stahovich, Randall Davis, PUI, 2001.
Summary
This paper discussed a system that was implemented for processing freehand sketching in an attempt to provide a method for natural interactions with a user interface. The interpretation of the freehand gestures into geometric descriptions was discussed, for a representation that can be used by the system easier. However, the interpretation part that was discussed within this paper is intended to be only the first part of a larger system that can provide understanding and interaction using freehand sketching.
Freehand sketching is more complicated than working off of a set of predefined shapes, since anything can be sketched but it still must be able to recognized by the system. Therefore, preprocessing is used in order to distinguish corners from curves to recognize the low-level geometric properties of the gesture. The processing stage includes three phases: approximation, beautification, and basic recognition. Approximation includes finding the vertices at the corners of the gesture by using a hybrid fit with both the stroke information and the timing information associated with sketching the gesture. It also consists of determining the curved sections of the gesture. The approximated data is then used within the beautification phase to improve the appearance of the gesture. The beautified data is used in the basic recognition phase to recognize basic geometric properties from the data.
The system was evaluated using a user study in which participants sketched a set of gestures using the system. Results from the evaluation labeled the system as easy and natural to use due to the ability to draw freehand gestures. In addition, it was determined that the system could efficiently and correctly interpret the freehand shapes that were drawn.
Thoughts
The intention of providing a system for allowing users to apply freehand sketching within user interfaces is an appealing idea. Something like this would open up a number of different interactions with user interfaces that have never been possible before. Since this paper simply described a single part in the process of creating such a system, it would be interesting to find out more about other parts of the system.
One of the most notable findings mentioned within this paper is the fact that timing data can be used to interpret gestures. In particular, it was explained that a user slows down when drawing corners in gestures, allowing corners to be recognized by timing data. One distracting thing that I noticed throughout the paper was that the usage of bounds and thresholds was mentioned multiple times but not how the values of those bounds and thresholds were actually determined. Nevertheless, the contributions of this paper regarding the recognition of freehand sketches seem to have been very important to the field of sketch recognition.
Title: Sketch Based Interfaces: Early Processing for Sketch Understanding
Authors: Tevfik Metin Sezgin, Thomas Stahovich, Randall Davis
Citation: "Sketch Based Interfaces: Early Processing for Sketch Understanding", Tevfik Metin Sezgin, Thomas Stahovich, Randall Davis, PUI, 2001.
Summary
This paper discussed a system that was implemented for processing freehand sketching in an attempt to provide a method for natural interactions with a user interface. The interpretation of the freehand gestures into geometric descriptions was discussed, for a representation that can be used by the system easier. However, the interpretation part that was discussed within this paper is intended to be only the first part of a larger system that can provide understanding and interaction using freehand sketching.
Freehand sketching is more complicated than working off of a set of predefined shapes, since anything can be sketched but it still must be able to recognized by the system. Therefore, preprocessing is used in order to distinguish corners from curves to recognize the low-level geometric properties of the gesture. The processing stage includes three phases: approximation, beautification, and basic recognition. Approximation includes finding the vertices at the corners of the gesture by using a hybrid fit with both the stroke information and the timing information associated with sketching the gesture. It also consists of determining the curved sections of the gesture. The approximated data is then used within the beautification phase to improve the appearance of the gesture. The beautified data is used in the basic recognition phase to recognize basic geometric properties from the data.
The system was evaluated using a user study in which participants sketched a set of gestures using the system. Results from the evaluation labeled the system as easy and natural to use due to the ability to draw freehand gestures. In addition, it was determined that the system could efficiently and correctly interpret the freehand shapes that were drawn.
Thoughts
The intention of providing a system for allowing users to apply freehand sketching within user interfaces is an appealing idea. Something like this would open up a number of different interactions with user interfaces that have never been possible before. Since this paper simply described a single part in the process of creating such a system, it would be interesting to find out more about other parts of the system.
One of the most notable findings mentioned within this paper is the fact that timing data can be used to interpret gestures. In particular, it was explained that a user slows down when drawing corners in gestures, allowing corners to be recognized by timing data. One distracting thing that I noticed throughout the paper was that the usage of bounds and thresholds was mentioned multiple times but not how the values of those bounds and thresholds were actually determined. Nevertheless, the contributions of this paper regarding the recognition of freehand sketches seem to have been very important to the field of sketch recognition.
Subscribe to:
Comments (Atom)