Sequence Learning Algorithm
A Sequence Learning Algorithm is a structured learning algorithm that can by implemented by a sequence learning system (to solve a sequence learning task).
- AKA: Sequential Data Training Method.
- Context:
- It can range from being a Supervised Sequence Labeling Algorithm, to being a Supervised Sequence Ranking Algorithm, to being a Supervised Sequence Estimation Algorithm.
- It can range from being a Supervised Sequence Learning Algorithm, to being a Semi-Supervised Learning Algorithm, to being a Unsupervised Sequence Learning Algorithm.
- Example(s):
- a Sequential Decision-Making Algorithm,
- a Sequence Labeling Algorithm,
- a Sequence Prediction Algorithm,
- a Sequence Generation Algorithm,
- a Sequence Recognition Algorithm,
- a Neural Sequence Learning Algorithm,
- a Hidden Markov Model Training Algorithm (for HMMs).
- a Conditional Random Field Training Algorithm (for CRFs).
- a Recurrent Neural Network Training Algorithm (for RNNs).
- a Sequence Labeling Algorithm.
- …
- Counter-Example(s):
- See: Language Model Learning Algorithm.
References
2013a
- (Wikipedia, 2013) ⇒ http://en.wikipedia.org/wiki/Sequence_learning#Sequence_learning_problems
- Sequence learning problems are used to better understand the different types of sequence learning. There are four basic sequence learning problems: sequence prediction, sequence generation, sequence recognition, and sequential decision making. These “problems” show how sequences are formulated. They show the patterns sequences follow and how these different sequence learning problems are related to each other.
Sequence prediction attempts to predict the next immediate element of a sequence based on all of the preceding elements. Sequence generation is basically the same as sequence prediction: an attempt to piece together a sequence one by one the way it naturally occurs. Sequence recognition takes certain criteria and determines whether or not the sequence is legitimate. Sequential decision making or sequence generation through actions breaks down into three variations: goal-oriented, trajectory-oriented, and reinforcement-maximizing. These three variations all want to pick the action(s) or step(s) that will lead to the goal in the future.[1]
These sequence learning problems reflect hierarchical organization of plans because each element in the sequences builds on the previous elements.
In a classic experiment published in 1967, Alfred L. Yarbus demonstrated that though subjects viewing portraits reported apprehending the portrait as a whole, their eye movements successively fixated on the most informative parts of the image. These observations suggest that underlying an apparently parallel process of face perception, a serial oculomotor process is concealed.[2] It is a common observation that when a skill is being acquired, we are more attentive in the initial phase, but after repeated practice, the skill becomes nearly automatic;[3] this is also known as unconscious competence. We can then concentrate on learning a new action while performing previously learned actions skillfully. Thus it appears that a neural code or representation for the learned skill is created in our brain, which is usually called procedural memory. The procedural memory encodes procedures or algorithms rather than facts.
- Sequence learning problems are used to better understand the different types of sequence learning. There are four basic sequence learning problems: sequence prediction, sequence generation, sequence recognition, and sequential decision making. These “problems” show how sequences are formulated. They show the patterns sequences follow and how these different sequence learning problems are related to each other.
2013b
- (Wikipedia, 2013) ⇒ http://en.wikipedia.org/wiki/Sequence_learning
- Sequence learning is inherent to human ability because it is an integrated part of conscious and nonconscious learning as well as activities. Sequences of information or sequences of actions are used in various everyday tasks: "from sequencing sounds in speech, to sequencing movements in typing or playing instruments, to sequencing actions in driving an automobile."[4] Sequence learning can be used to study skill acquisition and in studies of various groups ranging from neuropsychological patients to infants. According to Ritter and Nerb, “The order in which material is presented can strongly influence what is learned, how fast performance increases, and sometimes even whether the material is learned at all.”[5] Sequence learning, more known and understood as a form of explicit learning, is now also being studied as a form of implicit learning as well as other forms of learning. Sequence learning can also be referred to as sequential behavior, behavior sequencing, and serial order in behavior.
- ↑ Sun, Ron. "Introduction to Sequence Learning". http://www.cogsci.rpi.edu/~rsun/sun.seq-intro.ps. Retrieved 30 June 2011.
- ↑ Yarbus, Alfred L., "Eye movements during perception of complex objects", Yarbus, Alfred L., tr. Basil Haigh, ed. Lorrin A. Riggs, Eye Movements and Vision, New York: Plenum, 1967, OCLC 220267263, ch. 7, pp. 171–96.
- ↑ Fitts, P. M., "Perceptual motor skill learning", in Arthur W. Melton (ed.), Categories of Human Learning, New York: Academic Press, 1964, OCLC 180195, pp. 243–85.
- ↑ Clegg, Benjamin A; DiGirolamo, Gregory J, Keele, Steven W (August 1998). "Sequence learning". Trends in Cognitive Sciences 2 (8): 275–81. doi:10.1016/S1364-6613(98)01202-9.
- ↑ Frank E. Ritter et al., ed. (2007). In order to learn: how the sequence of topics influences learning. Oxford series on cognitive models and architectures. Oxford/New York: Oxford University Press. ISBN 978-0-19-517884-5.
2006
- (Bishop, 2006) ⇒ Christopher M. Bishop. (2006). “Pattern Recognition and Machine Learning. Springer, Information Science and Statistics.
2001
- (Sun, 2001) ⇒ Ron Sun. (2001). “Introduction to Sequence Learning.” In: Sequence Learning - Paradigms, Algorithms, and Applications. ISBN:978-3-540-44565-4 doi:10.1007/3-540-44565-X_1
- QUOTE: ... With some necessary simplification, we can categorize various sequence learning problems that have been tackled into the following categories: (1) sequence prediction, in which, we want to predict elements of a sequence based on the preceding element(s): (2) sequence generation, in which we want to generate elements of a sequence one by one in their natural order: and (3) sequence recognition, in which we want to determine if a sequence is a legitimate one according to some criteria: in addition, (4) sequential decision making involves selecting sequence of actions, to accomplish a goal, to follow a trajectory, or to maximize (or minimize) a reinforcement (or cost) function that is normally the (discounted) sum of reinforcements (costs) that are received along the way (see Bellman 1957, Bertsekas and Tsitsiklis 1995) ...