Showing papers on "Sequence learning published in 1995"

Is Learning The n-th Thing Any Easier Than Learning The First?

Abstract: This paper investigates learning in a lifelong context. Lifelong learning addresses situations in which a learner faces a whole stream of learning tasks. Such scenarios provide the opportunity to transfer knowledge across multiple learning tasks, in order to generalize more accurately from less training data. In this paper, several different approaches to lifelong learning are described, and applied in an object recognition domain. It is shown that across the board, lifelong learning approaches generalize consistently more accurately from less training data, by their ability to transfer knowledge across learning tasks.

Learning in Humans and Machines : Towards an Interdisciplinary Learning Science

Abstract: Chapter headings: Towards an Interdisciplinary Learning Science (P. Reimann, H. Spada). A Cognitive Psychological Approach to Learning (S. Vosniadou). Learning to Do and Learning to Understand: A Lesson and a Challenge for Cognitive Modeling (S. Ohlsson). Machine Learning: Case Studies of an Interdisciplinary Approach (W. Emde). Mental and Physical Artifacts in Cognitive Practices (R. Saljo). Learning Theory and Instructional Science (E. De Corte). Knowledge Representation Changes in Humans and Machines (L. Saitta and Task Force 1). Multi-Objective Learning with Multiple Representations (M. Van Someren, P. Reimann). Order Effects in Incremental Learning (P. Langley). Situated Learning and Transfer (H. Gruber et al.). The Evolution of Research on Collaborative Learning (P. Dillenbourg et al.). A Developmental Case Study on Sequential Learning: The Day-Night Cycle (K. Morik, S. Vosniadou). Subject index. Author index.

The Psychology of Associative Learning

Abstract: It is hard to think of any significant aspect of our lives that is not influenced by what we have learned in the past. Of fundamental importance is our ability to learn the ways in which events are related to one another, called associative learning. This book provides a fresh look at associative learning theory and reviews extensively the advances made over the last twenty years. The Psychology of Associative Learning begins by establishing that the human associative learning system is rational in the sense that it accurately represents event relationships. David Shanks goes on to consider the informational basis of learning, in terms of the memorisation of instances, and discusses at length the application of connectionist models to human learning. The book concludes with an evaluation of the role of rule induction in associative learning. This will be essential reading for graduate students and final year undergraduates of psychology.

Closing the Gap Between Learning and Instruction

Abstract: In this paper, I explore the mismatch between the pedagogical intentions and plans of the educational institution, curriculum, teacher, and textbook, and the outcomes as realized through the skills and knowledge that learners take away from instructional encounters. Although there will never be a one-to-one relationship between teaching and learning, there are ways in which teachers and learners and teaching and learning can be brought closer together. In this article, I look at ways of closing the gap in relation to experiential content, learning process, and language content. The theme holding these three disparate domains together is that of learner centredness, and it is this concept which I take as my point of departure.

A PET Investigation of implicit and explicit sequence learning

Abstract: The purpose of this study was to determine the mediating neuroanatomy of implicit and explicit sequence learning using a modified version of the serial reaction time (SRT) paradigm. Subjects were seven healthy, right-handed adults (three male, four female, mean age 26.7, range 18–43 yr). PET data were acquired via the oxygen-15-labeled-carbon dioxide inhalation method while subjects performed the SRT. Subjects were scanned during two blocks each of (1) no sequence (Random), (2) single-blind, 12-item sequence (Implicit), and (3) unblinded, same sequence (Explicit). Whole-brain-normalized images reflecting relative regional cerebral blood flow (rCBF) were transformed to Talairach space, and statistical parametric maps (SPMs) of z-scores were generated for comparisons of interest. The threshold for significant activation was defined as z-score ≥ 3.00. Behavioral data demonstrated significant learning (P < .05) for Implicit and Explicit conditions. Tests of explicit knowledge reflected non-significant explicit contamination during the Implicit condition. Foci of significant activation in the Implicit condition were found in right ventral premotor cortex, right ventral caudate/nucleus accumbens, right thalamus, and bilateral area 19; activation in the Explicit condition included primary visual cortex, peri-sylvian cortex, and cerebellar vermis. Activations in visual and language areas during the Explicit condition may reflect conscious learning strategies including covert verbal rehearsal and visual imagery. Right-sided premotor, striatal, and thalamic activations support the notion that implicit sequence learning is mediated by cortico-striatal pathways, preferentially within the right hemisphere. © 1996 Wiley-Liss, Inc.

On the Modularity of Sequence Representation

Abstract: A modular theory of motor control posits that the representation of an action sequence is independent of the effector (motor) system that implements the sequence. Three experiments tested this theory. Each used a variant of a method developed by Nissen and Bullemer (1987) in which subjects responded to visual signals occupying different spatial positions by pressing a key corresponding to each signal position. Sequence learning is indicated when reaction times to signals that follow a sequence become faster with practice than reaction times to random signals. The first experiment showed transfer of sequential learning of key pressing from the fingers to the arms, or vice versa. Similar transfer was found when a distraction task was added that likely blocked an attentional form of learning (cf. Curran & Keele, 1993). In a third experiment, much but not all of the sequential learning transferred from a situation in which the response was a key press to one with a vocal response, suggesting that at ...

A model of corticostriatal plasticity for learning oculomotor associations and sequences

Abstract: We present models that learn context-dependent oculomotor behavior in (1) conditional visual discrimination and (2) sequence reproduction tasks, based on the following three principles: (1) Visual input and efferent copies of motor output produce patterns of activity in cortex. (2) Cortex influences the saccade system in part via corticostriatal projections. (3) A reinforcement learning mechanism modifies corticostriatal synapses to link patterns of cortical activity to the correct saccade responses during trial-and-error learning. Our conditional visual discrimination model learns to associate visual cues with the corresponding saccades to one of two left-right targets. A visual cue produces patterns of neuronal activity in inferotemporal cortex (IT) that projects to the oculomotor region of the striatum. Initially random saccadic "guesses," when directed to the correct target for the current cue, result in increased synaptic strength between the cue-related IT cells and the striatal cells that participate in the correct saccade, increasing the probability that this cue will later elicit the correct saccade. We show that the model generates "inhibitory gradients" on the striatum as the substrate for spatial generalization. Our sequence reproduction model learns, when presented with temporal sequences of spatial targets, to reproduce the corresponding sequence of saccades. At any point in the execution of a saccade sequence, the current pattern of activity in pre-frontal cortex (PFC), combined with visual input and the motor efferent copy of the previous saccade, produces a new pattern of activity in PFC to be associated with the next saccade. Like IT, PFC also projects to the oculomotor region of the striatum. Correct guesses for the subsequent saccade in the sequence results in strengthening of corticostriatal synapses between active PFC cells and striatal cells involved in the correct saccade. The sequence is thus reproduced as a concatenation of associations. We compare the results of this model with data previously obtained in the monkey and discuss the nature of cortical representations of spatiotemporal information.

Applications of inductive logic programming

Abstract: Techniques of machine learning have been successfully applied to various problems [1, 12]. Most of these applications rely on attribute-based learning, exemplified by the induction of decision trees as in the program C4.5 [20]. Broadly speaking, attribute-based learning also includes such approaches to learning as neural networks and nearest neighbor techniques. The advantages of attribute-based learning are: relative simplicity, efficiency, and existence of effective techniques for handling noisy data. However, attribute-based learning is limited to non-relational descriptions of objects in the sense that the learned descriptions do not specify relations among the objects' parts. Attribute-based learning thus has two strong limitations: the background knowledge can be expressed in rather limited form, andthe lack of relations makes the concept description language inappropriate for some domains.

Complex sensory-motor sequence learning based on recurrent state representation and reinforcement learning

Abstract: A novel neural network model is presented that learns by trial-and-error to reproduce complex sensory-motor sequences. One subnetwork, corresponding to the prefrontal cortex (PFC), is responsible for generating unique patterns of activity that represent the continuous state of sequence execution. A second subnetwork, corresponding to the striatum, associates these state-encoding patterns with the correct response at each point in the sequence execution. From a neuroscience perspective, the model is based on the known cortical and subcortical anatomy of the primate oculomotor system. From a theoretical perspective, the architecture is similar to that of a finite automaton in which outputs and state transitions are generated as a function of inputs and the current state. Simulation results for complex sequence reproduction and sequence discrimination are presented.

A network approach to probabilistic information retrieval

Abstract: In this article we show how probabilistic information retrieval based on document components may be implemented as a feedforward (feedbackward) artificial neural network. The network supports adaptation of connection weights as well as the growing of new edges between queries and terms based on user relevance feedback data for training, and it reflects query modification and expansion in information retrieval. A learning rule is applied that can also be viewed as supporting sequential learning using a harmonic sequence learning rate. Experimental results with four standard small collections and a large Wall Street Journal collection (173,219 documents) show that performance of feedback improves substantially over no feedback, and further gains are obtained when queries are expanded with terms from the feedback documents. The effect is much more pronounced in small collections than in the large collection. Query expansion may be considered as a tool for both precision and recall enhancement. In particular, small query expansion levels of about 30 terms can achieve most of the gains at the low-recall high-precision region, while larger expansion levels continue to provide gains at the high-recall low-precision region of a precision recall curve.

A dual-phase model for the individual learning process in industrial tasks

Abstract: A new dual-phase model for learning industrial tasks is presented, based on the combined effects of cognitive and motor processes. The model proposes that cognitive elements dominate learning during the early cycles, whereas motor elements dominate the learning process as the number of repetitions becomes large. The implication is that the observed learning slope is a variable whose value gradually increases as experience is gained. Experimental studies are described whose results support the behavior of the dual-phase learning model.

When encoding fails: instructions, feedback, and registration without learning.

Abstract: Four experiments replicated and extended the registration-without-learning effect, in which there is little improvement in the ability to discriminate an old target (X) from a highly similar test item (Y) after the first few presentations of X, even though judgments of frequency continue to rise in an openended fashion. Forced-choice testing revealed the anomalous form of the learning curve for X-Y discrimination (faster and then slower than the exponential). Effects of several different learning instructions were compared, but these appeared to affect only the level of initial learning, and to do little to promote X-Y discrimination learning on later presentations. The opportunity for self-testing with feedback during study provided no benefits when responding was covert, but did when overt anticipation was required. The findings are discussed in relation to the roles of bottom-up and topdown processing in memory encoding, and to the importance of error-correcting feedback in further structural learning of materials, once the materials have become familiar.

DNA Sequence Classification Using Compression-Based Induction

Abstract: Inductive learning methods, such as neural networks and decision trees, have become a popular approach to developing DNA sequence identification tools. Such methods attempt to form models of a collection of training data that can be used to predict future data accurately. The common approach to using such methods on DNA sequence identification problems forms models that depend on the {\em absolute locations} of nucleotides and assume {\em independence} of consecutive nucleotide locations. This paper describes a new class of learning methods, called {\em compression-based induction} (CBI), that is geared towards sequence learning problems such as those that arise when learning DNA sequences. The central idea is to use text compression techniques on DNA sequences as the means for generalizing

Effects of learning contexts on implicit and explicit learning.

Abstract: Two parallel tasks involving rule learning were identified in Experiment 1A and were used to assess implicit and explicit learning. In both tasks, subjects had to input numbers in order to reach the target values of outputs. The relationship between inputs and outputs was either simple (in the simple task) or complex (in the complex task), and the way in which target values were presented could be in the form of either numbers (in the simple task) or lines (in the complex task). Experiment 1B examined the validity of the explicit measure in the complex task. Experiments 2–4 investigated the interaction between learning contexts and the simple/complex learning tasks. Verbalization and instructions to search for the rules facilitated the simple-task learning and hurt or have no effect on the complex-task learning. In the observational-learning condition, no learning occurred for the simple task, and the complex task learning was impaired. These results suggest that the complex task and simple task involve two distinct learning systems. Other implications are also discussed.

Levels in Learning

Abstract: In 1957, Van Hiele published his study on levels in learning mathematics. In this article we present an elaboration of the theory on levels in learning and an illustration of this theory based on empirical results. We distinguish between three levels: The first is the formation of an image out of a range of familiar examples or experiences. Second, on the basis of the image a schema can be built, which includes all kinds of interrelated details. Third, a theory can be developed with basic assumptions, definitions, and logical inferences. Before entering the next level, the learner should be sufficiently experienced on the actual level. It is shown that the theory on levels in learning can be integrated with other learning principles, such as notions from constructivism.

Learning to reason with a restricted view

Abstract: The Learning to Reason framework combines the study of Learning and Reasoning into a single task. Within it, learning is done specifically for the purpose of reasoning with the learned knowledge. Computational considerations show that this is a useful paradigm; in some cases learning and reasoning problems that are intractable when studied separately become tractable when performed as a task of Learning to Reason.

Indirect learning of event sequences : the effects of divided attention and stimulus continuity

Abstract: In a serial reaction time (SRT) task, the learning curve is sleeper when the stimuli are presented in a repeating sequential manner rather than in random order (Nissen & Bullemer, 1987). This is true even when subjects report being unaware of the presence of the repeating sequence. The present study examines the nature of this learning under conditions designed to reduce attentional resources and to disrupt the continuity of stimuli. In the first three experiments, subjects were trained in the SRT task, with or without the addition of a secondary tone counting task, and with repeating or non-repeating sequences. The results suggest that some sequence learning occurred despite the presence of a secondary task. Experiment 4 examined the extent of sequence learning when the inter-stimulus interval was varied between trials. The overall results suggest that despite reduced attentional allocation and discontinuous stimulus presentation, some sequence learning occurs. This result supports other work suggesting a dissociation between learning when measured explicitly, and when assessed through performance indicators.

The Psychology of Learning: Principles and Processes

Abstract: 1. Introduction. 2. Classical Conditioning. 3. Instrumental Learning. 4. Operant Conditioning. 5. The Nature of Reinforcement. 6. Motor Learning. 7. Language Acquisition. 8. Verbal Learning. 9. Memory. 10. Relationships, Concepts, and Thinking. 11. Neural Processes in Learning. 12. Behavior Therapy. References. Index.

Analogical Transfer in Sequence Learning.

Abstract: Analogical transfer in problem solving is a fundamental aspect of human intelligence that involves exploiting knowledge about the solution for one problem to solve an0ther.l In order to provide a quantifiable measure of analogical transfer in sequence learning (ATSL), we developed a novel extension of the serial reaction time (SRT) paradigm. To quantify the underlying representational and computational requirements for ATSL we extended a neurobiologically based model of primate prefrontal cortex (PFC) and basal ganglia function in visuomotor sequence learning* to address analogical transfer. We compared the behavior of this model with preliminary data from normal subjects and patients with frontostriatal dysfunction [idiopathic dopa-sensitive Parkinson’s disease (PD)] in the ATSL paradigm.

Implicit learning and concept learning.

Abstract: In Experiments 1 and 2, subjects were exposed to letter strings that followed a pattern--the second letter was always the same. This exposure was disguised as a test of immediate memory. Following this training, subjects could discriminate new letter strings following the pattern from letter strings not following the pattern more often than would be expected by chance, which is the traditional evidence for concept learning. Discrimination was also better than would be predicted from subjects' explicit report of the pattern, demonstrating the co-occurrence of concept learning and implicit learning. In Experiment 3, rules were learned explicitly. Discrimination was worse than would be predicted from subjects' explicit report, validating the implicit learning paradigm. In Experiment 4, deviations from a prototypical pattern were presented during training. In the test of discrimination, prototypes were as familiar as old deviations and more familiar than new deviations, even when considering only implicit knowledge. Experiment 5 found implicit knowledge of a familiar concept. These results are consistent with the hypothesis that the distinguishing features of a concept can be learned implicitly, and that one type of implicit learning is concept learning.

Modularity of Sequence Learning Systems in Humans

Abstract: Humans excel at a variety of learned and highly skilled activities in which complex sequential behavior is distributed over time. The major theme of this chapter concerns the hypothesis that sequence learning and production of sequences of activities involves not a single function, but rather is made up of multiple components. For example, in playing a piano, pitch is mapped to key position and key position is mapped to the motor system for bringing the arms, hands, and fingers to the keys. In addition to this spatial mapping, the pianist must learn the sequence of notes or keys that correspond to a piece of music. The sequential representation must indicate not only which note or key is next in a series, but must also specify the intervals at which the keys should be hit and with what intensity. In other activities, dancing for example, trajectory through space, and not just the target of movement, must be specified. It is likely that some of these functions are independent of one another, both in the psychological sense that one function can be affected with minimal or no influence on another, and in a neurobiological sense in that they depend on different brain regions. This chapter will focus on a selected aspect of skill, the representation of learned sequences, and will consider only those representations that specify the succession of events. One of the issues to be addressed is the relationship between the representation of a sequence and the motor system that actually produces the sequence. Evidence will be presented that sequence representation is relatively abstract and independent of the implementation system. A second line of evidence to be presented suggests that the sequential representation itself has constituent parts or modules.

Learning control method

Abstract: PURPOSE:To highly precisely predict a physical phenomenon in the work of the next material by performing updating separately for a learning coefficient for respective layers corresponding to the respective layers and a time sequential learning coefficient in common to the respective layers at the time of correcting a control model CONSTITUTION:In the case of updating the learning coefficient, a control result and a control target value are compared (step ST12) and when the difference is within a predetermined threshold value, the learning coefficient is not update When the difference exceeds the threshold value, the learning coefficient CSo(i) for the respective layers is read from the instantaneous value C of the learning coefficient, and a table and the instantaneous value CT of the times sequential learning coefficient is calculate (ST13) Then, the time sequential learning coefficient CTo before the updating is read from the table and the new time sequential learning coefficient CTn is calculated (ST14) Then, by using the learning coefficient CTn, the instantaneous value CS(i) of the learning coefficient for the respective layers is calculated (ST15) Then, from the learning coefficient CSo(i) before the updating and the instantaneous value CS(i), the new learning coefficient CS(i) for the respective layers is calculate (ST16)

Effective Techniques for Teaching Highly Gifted Visual-Spatial Learners.

Abstract: This paper discusses the differences between spatial and sequential learning, characteristics of individuals who exhibit stronger visual-spatial learning, and strategies for teaching children with visual-spatial strengths. Techniques include: (1) using visual aids, such as overhead projectors, and visual imagery in lectures; (2) using manipulative materials to allow hands-on experience; (3) using a sight approach to reading rather than phonics; (4) using a visualization approach to spelling; (5) having students discover their own methods of problem-solving; (6) avoiding rote memorization; (7) avoiding drill and repetition; (8) finding out what students have already mastered before teaching them; (9) giving students advanced material at a faster pace; (10) allowing students to accelerate in school; (11) emphasizing mastery of higher level concepts; (12) emphasizing creativity, imagination, new insights, and new approaches rather than acquisition of knowledge; (13) grouping gifted visual-spatial learners together for instruction; (14) engaging students in independent studies for group projects which involve problem-finding as well as problem-solving; (15) allowing students to construct, draw, or otherwise create visual representations of concepts; (16) using computers so that material is presented visually; and (17) having the students discuss the ethical, moral, and global implications of their learning and involving them in service-oriented projects. (CR) Reproductions supplied by EDRS are the best that can be made from the original document. PERMISSION TO REPRODUCE AND DISSEMINATE THIS MATERIAL HAS BEEN GRANTED BY QA/MaIr TO THE EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) EFFECTIVE TECHNIQUES FOR TEACHING HIGHLY GIFTED VISUAL-SPATIAL LEARNERS Linda Kreger Silverman, Ph.D. Gifted Development Center Denver, Colorado U.S. DEPARTMENT OF EDUCATION Office of Educational Research and Improvement EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) This document has been reproduced as received from the person or organization originating it. Minor changes have been made to improve reproduction quality. Points of view or opinions stated in this document do not necessarily represent official OERI position or policy. Spatial and sequential dominance are two different mental organizations that affect perceptions and apparently lead to different world views. Information deemed central to one viewpoint appears irrelevant from the other perspective. The sequential system appears to be profoundly influenced by audition, whereas the spatial system relies heavily on vision and visualization. Auditory-sequential learners are extremely aware of time but may be less aware of space; visual-spatial learners are often preoccupied with space at the expense of time. Sequential learning involves analysis, orderly progression of knowledge from simple to complex, skillful categorization and organization of information, and linear, deductive reasoning. Spatial learning involves synthesis, intuitive grasp of complex systems (skipping many of the foundational "steps"), simultaneous processing of concepts, inductive reasoning, active use of imagery, and idea generation by combining disparate elements in new ways. These diverse ways of relating to the world have had powerful ramifications throughout history in the development of various philosophies, religions, cultures, branches of science, and psychological theories. Western and Eastern philosophies and cultures provide dramatic examples of these differences. Western thought is sequential, temporal, analytic; Eastern thought is spatial and holistic (Bolen, 1979). Cause and effect sequences are stressed in Euro-American ideation, whereas synchronicity of unrelated events is appreciated from an Asian world view. Western languages are constructed out of non-meaningful elements--letters of the alphabet; Eastern languages traditionally have been composed of pictorial representations. Perhaps the greater facility of Asian children in the visual-spatial domain can be traced at least in part to the emphasis on visualization in the linguistic system. Temporal, sequential and analytical functions are thought to be left-hemispheric strengths, while spatial, holistic and synthetic functions are considered right-hemispheric strengths (Dixon, 1983; Gazzaniga, 1992; Springer & Deutsch, 1989; West, 1991). However, most researchers agree that integration of both hemispheres is necessary for higher-level thought processes. We all use both hemispheres, but not with equal facility. Highly gifted individuals show strong integration of sequential and spatial functions, but most of the gifted children we have assessed seem naturally to favor one or the other mode. These different mental organizations appear to be innate. Although one can gain more facility with one or the other mode through learning, it is unlikely that a person with sequential dominance can learn to perceive the world in exactly the same way as an individual with spatial dominance or vice versa. Instead of trying to remake one or the other style of learning, we need to accept these inherent differences in perception, and appreciate their complementarity since we inhabit a spatial-temporal reality. When these differences are not understood, there is dissension; when they are honored, they enable an exchange of information that forms a more complete conception of reality than can be gained by either perspective in isolation.

ILA: Combining Inductive Learning with Prior Knowledge and Reasoning

Abstract: Much effort has been devoted to understanding learning and reasoning in artificial intelligence. However, very few models attempt to integrate these two complementary processes. Rather, there is a vast body of research in machine learning, often focusing on inductive learning from examples, quite isolated from the work on reasoning in artificial intelligence. Though these two processes may be different, they are very much interrelated. The ability to reason about a domain of knowledge is often based on rules about that domain, that must be learned somehow. And the ability to reason can often be used to acquire new knowledge, or learn. This paper introduces an Incremental Learning Algorithm (ILA) that attempts to combine inductive learning with prior knowledge and reasoning. ILA has many important characteristics useful for such a combination, including: 1) incremental, self-organizing learning, 2) non-uniform learning, 3) inherent non-monotonicity, 4) extensional and intensional capabilities, and 5) low order polynomial complexity. The paper describes ILA, gives simulation results for several applications, and discusses e ach of the above characteristics in detail.

Practical PAC learning

Abstract: We present new strategies for "probably approximately correct" (par) learning that use fewer training examples than previous approaches. The idea is to observe training examples one-at-a-time and decide "on-line" when to return a hypothesis, rather than collect a large fixed-size training sample. This yields sequential learning procedures that par-learn by observing a small random number of examples. We provide theoretical bounds on the expected training sample size of our procedure -- but establish its efficiency primarily by a scries of experiments which show sequential learning actually uses many times fewer training examples in practice. These results demonstrate that pac-learning can be far more efficiently achieved in practice than previously thought.

Implicit learning in the presence of multiple cues

Abstract: Is implicit learning an independent and automatic process? In this paper, I attempt to answer this question by exploring whether implicit learning occurs even despite the availability of more reliable explicit information about the material to be learnt. I report on a series of experiments during which subjects performed a sequential choice reaction task. On each trial subjects were exposed to a stimulus and to a cue of varying validity which, when valid, indicated where the next stimulus would appear. Subjects could therefore optimize their performance either by implicitly encoding the sequential constraints contained in the material or by explicitly relying on the information conveyed by the cue. Some theories predict that implicit learning does not rely on the same processing resources as involved in explicit learning. Such theories would thus predict that sensitivity to sequential constraints should not be affected by the presence of reliable explicit information about sequence structure. Other theories, by contrast, would predict that implicit learning would not occur in such cases. The results suggest that the former theories are correct. I also describe preliminary simulation work meant to enable the implications of these contrasting theories to be explored.

Learning at the crossroads of biology and computation

Abstract: Discusses various avenues for exploiting biological learning mechanisms within machine learning. Special attention is given to the following issues: (a) the reasons for the wide variety of biological learning mechanisms; (b) the relation between lifetime and genetic learning; (c) a description of the driving forces of genetic learning and their use in evolutionary computation. Various symbolic machine learning and reasoning techniques can be used to complement (genetic and/or neural) sub-symbolic learning. A first approach uses symbolic induction for explaining the behavior of (genetically evolved) neural nets. Next, a general framework for the use of (symbolic) domain knowledge during genetic learning is introduced. >