Home | Students | Course Content | Homework | Projects |Announcements

Homework 2: how to attain expertise

1. BASIC LANGUAGE: here your instructions are very basic, e.g. "move hand by dX dY dZ", or "if string is in-between, close index finger on thumb", or "press the left-arrow-key until X", etc. give a few sample "programs" that you gave to the robot. please try to record the video of the robot "executing" these programs. what are the difficulties in defining a program for doing the task as efficiently as you can?
2. FUNCTIONS. you may now add functions that are defined on top of the basic language of step A. such a system may need to build up a hierarchy of actions, each of which will operate on a set of objects along with its own parameters, and execute a subtask. the higher the level of the function, the more it chunks a large number of possibilities into a single behavioural pattern. describe some functions from the top two levels in your hierarchy, and define a top level function in terms of the ones below it. check if the robot can execute this function, as defined.
3. EXPLICIT OR IMPLICIT? can you, as the human expert, describe exactly what you are doing so that it can be given as a program? is the chess expert's knowledge explicit? relate the explicitness of knowledge here to that of the chess expert.
4. LEARNING: as an alternative, would it be better to design a robot that can learn and improve its performance over repeated training episodes? consider how it could a) decompose the task into the functions as in B, and b) discuss the learning of a particular function from repeated trials. which would serve this learning better - a symbolic or a connectionist model?
5. EMBODIMENT: consider a pure mind - i.e. a computer without a body. such a system may be able to execute a program for the motions, but can it learn the motions? consider your answer to C.
6. PARAMETERS: you may do the same task under other kinds of constraints (e.g. play a video game that is frequently interrupted); you may do it more rapidly or more slowly; more accurately; in the dark; in order to explore more of the alternatives rather than just score higher; etc. consider how the such variants may be incorporated into the functions in B in terms of parameters.
7. REPRESENTATION and DIMENSIONALITY: it is believed that expertise is a matter of improved representation of the problem in terms of chunks. consider the the high-dimensional space in which the solutions to the task - e.g. the optimal sequence of motions - lie. estimate the dimensionality of this space assuming the action is executed over 100 time steps. once the task has been mapped to the set of functions in B or F, how does it change the dimensionality?
8. RULES OR PROCEDURES? when teaching arithmetic to a computer, you are better off teaching it as rules or by the method followed above? discuss the differences between the tasks given above and the tasks in arithmetic, due to which these areas are different. why is learning math harder for humans than for machines? consider the ramifications of this process in terms of designing education programs for teaching kids mathematics.
9. SYMBOLIC vs CONNECTIONIST: Explain how a symbolic model may map this task. How would it differ if we were to use a connectionist model to approach this task? What do you think more closely resembles the processes in your mind?

Groups

Submissions