VS265: Class project - Revision history

Bruno at 04:13, 20 November 2014

2014-11-20T04:13:05Z

← Older revision		Revision as of 04:05, 20 November 2014
Line 1:		Line 1:
	'''Project presentations''' will take place on project presentation day ~~(TBD)~~. You may present your project either as an oral (15 min.) or poster presentation.		'''Project presentations''' will take place on project presentation day: Monday, Dec. 15. You may present your project either as an oral (15 min.) or poster presentation.

	Think of the class project as an extended lab assignment. In most of the labs we just scratched the surface of various network models. This is a chance to explore these ideas in more depth. In addition to giving an oral or poster presentation on your project at the end of the term, you should turn in a short writeup (5 pages) that describes the problem you investigated, why it is interesting, your approach, results, and conclusions.		Think of the class project as an extended lab assignment. In most of the labs we just scratched the surface of various network models. This is a chance to explore these ideas in more depth. In addition to giving an oral or poster presentation on your project at the end of the term, you should turn in a short writeup (5 pages) that describes the problem you investigated, why it is interesting, your approach, results, and conclusions.

Bruno: /* Data */

2014-10-24T04:15:50Z

Data

← Older revision		Revision as of 04:07, 24 October 2014
Line 58:		Line 58:
	* [http://www.cs.toronto.edu/~roweis/data.html '''Sam Roweis'''] - Many datasets in Matlab format: MNIST and USPS handwritten digits, faces, text, speech.		* [http://www.cs.toronto.edu/~roweis/data.html '''Sam Roweis'''] - Many datasets in Matlab format: MNIST and USPS handwritten digits, faces, text, speech.

	* '''Hans van Hateren''' - Natural stimuli collection: still images, intensity time series, video.		* '''Hans van Hateren''' - Natural stimuli collection: still images, intensity time series, video. [http://pirsquared.org/research/#van-hateren-database mirror]
	<!-- [http://www.kyb.mpg.de/bethge/vanhateren/index.php mirror 2](Germany, detailed description), -->~~[http://pirsquared.org/research/#van-hateren-database mirror]~~		<!-- [http://www.kyb.mpg.de/bethge/vanhateren/index.php mirror 2](Germany, detailed description), -->

	* [http://www.robjhyndman.com/TSDL/ '''Time Series Data Library''']		* [http://www.robjhyndman.com/TSDL/ '''Time Series Data Library''']

	* [http://www.face-rec.org/databases/ '''Face Recognition Databases''']		* [http://www.face-rec.org/databases/ '''Face Recognition Databases''']

Bruno: /* Data */

2014-10-24T04:15:07Z

Data

← Older revision		Revision as of 04:07, 24 October 2014
Line 58:		Line 58:
	* [http://www.cs.toronto.edu/~roweis/data.html '''Sam Roweis'''] - Many datasets in Matlab format: MNIST and USPS handwritten digits, faces, text, speech.		* [http://www.cs.toronto.edu/~roweis/data.html '''Sam Roweis'''] - Many datasets in Matlab format: MNIST and USPS handwritten digits, faces, text, speech.

	* ~~[http://hlab.phys.rug.nl/archive.html~~ '''Hans van Hateren'''] - Natural stimuli collection: still images, intensity time series, video.		* '''Hans van Hateren''' - Natural stimuli collection: still images, intensity time series, video.
	~~''Note:'' Hans van Hateren's website is down, there are two mirrors for the still images collection:~~ [http://www.kyb.mpg.de/bethge/vanhateren/index.php mirror 2](Germany, detailed description), [http://pirsquared.org/research/#van-hateren-database mirror 2] ~~(US,CA - faster)~~		<!-- [http://www.kyb.mpg.de/bethge/vanhateren/index.php mirror 2](Germany, detailed description), -->[http://pirsquared.org/research/#van-hateren-database mirror]

	* [http://www.robjhyndman.com/TSDL/ '''Time Series Data Library''']		* [http://www.robjhyndman.com/TSDL/ '''Time Series Data Library''']

	* [http://www.face-rec.org/databases/ '''Face Recognition Databases''']		* [http://www.face-rec.org/databases/ '''Face Recognition Databases''']

Bruno at 04:12, 24 October 2014

2014-10-24T04:12:59Z

← Older revision		Revision as of 04:04, 24 October 2014
Line 49:		Line 49:
	* '''Memristors.''' The memristor was originally proposed by Leon Chua at UC Berkeley back in the 1970's as a hypothetical '4th circuit element.' Recently, a group at [http://redwood.berkeley.edu/vs265/memristor-nature08.pdf HP labs] has discovered a memristive like element (see also [http://spectrum.ieee.org/semiconductors/processors/how-we-found-the-missing-memristor this article]). What is intriguing about the memristor is its synapse like properties ([http://redwood.berkeley.edu/vs265/memristor-stdp.pdf ref]), and according to Chua it even leads to a more straightforward model of action potential dynamics! Might memristive elements exist in neural circuits in the brain? Try simulating a memristive device and explore its relevance to plasticity or dynamics in neural systems. -->		* '''Memristors.''' The memristor was originally proposed by Leon Chua at UC Berkeley back in the 1970's as a hypothetical '4th circuit element.' Recently, a group at [http://redwood.berkeley.edu/vs265/memristor-nature08.pdf HP labs] has discovered a memristive like element (see also [http://spectrum.ieee.org/semiconductors/processors/how-we-found-the-missing-memristor this article]). What is intriguing about the memristor is its synapse like properties ([http://redwood.berkeley.edu/vs265/memristor-stdp.pdf ref]), and according to Chua it even leads to a more straightforward model of action potential dynamics! Might memristive elements exist in neural circuits in the brain? Try simulating a memristive device and explore its relevance to plasticity or dynamics in neural systems. -->

	* '''Learning in sensorimotor loops.''' A goal of research in both robotics and neuroscience is to understand the principles of adaptive behavior in embodied systems. However, currently there are few theories for guiding work in this area. One approach advanced by O'regan and colleagues is based on learning "sensorimotor contingencies" ([http://redwood.berkeley.edu/vs265/noe-oregan.pdf ref1], [http://redwood.berkeley.edu/vs265/philipona-oregan.pdf ref2]). Another by Ralf Der and colleagues is based on minimizing both predictive and 'postdictive' error ([http://redwood.berkeley.edu/vs265/taming_the_beast_martius2010.pdf ref]). Both propose simple algorithmic examples that you can implement in computer simulation, or you may ~~wish~~ to ~~implement in~~ a simple robotic platform.		* '''Learning in sensorimotor loops.''' A goal of research in both robotics and neuroscience is to understand the principles of adaptive behavior in embodied systems. However, currently there are few theories for guiding work in this area. One approach advanced by O'regan and colleagues is based on learning "sensorimotor contingencies" ([http://redwood.berkeley.edu/vs265/noe-oregan.pdf ref1], [http://redwood.berkeley.edu/vs265/philipona-oregan.pdf ref2]). Another by Ralf Der and colleagues is based on minimizing both predictive and 'postdictive' error ([http://redwood.berkeley.edu/vs265/taming_the_beast_martius2010.pdf ref]). Both propose simple algorithmic examples that you can implement in computer simulation, or you may want to try implementing on a simple robotic platform.

Bruno at 04:09, 24 October 2014

2014-10-24T04:09:57Z

← Older revision		Revision as of 04:01, 24 October 2014
Line 7:		Line 7:
	This is not an exhaustive listing, just some suggestions to get you started thinking.		This is not an exhaustive listing, just some suggestions to get you started thinking.

	* '''Dendritic nonlinearities.''' Most of the models we have discussed are based upon the ~~McCulloch-Pitts~~ model - a linear weighted sum of inputs followed by a single output nonlinearity (threshold or sigmoid). However as we saw in the first week of class, even the passive properties of dendrites are such that inputs combine nonlinearly. There are also active processes (e.g., dendritic spikes) that are highly nonlinear. This would seem to imply that the concept of linear separability, which is central to machine learning, is perhaps not a fundamental limitation for neural systems. How might you exploit the nonlinear properties of dendrites pattern discrimination and learning? What are the consequences of connecting such elements together in recurrent circuits? Do they still exhibit attractor dynamics? (See work of [http://lnc.usc.edu/publications.htm Bartlett Mel] to get started here, esp. papers from early 90's and early 2000's. )		* '''Dendritic nonlinearities.''' Most of the models we have discussed are based upon the Perceptron model - a linear weighted sum of inputs followed by a single output nonlinearity (threshold or sigmoid). However as we saw in the first week of class, even the passive properties of dendrites are such that inputs combine nonlinearly. There are also active processes (e.g., dendritic spikes) that are highly nonlinear. This would seem to imply that the concept of linear separability, which is central to machine learning, is perhaps not a fundamental limitation for neural systems. How might you exploit the nonlinear properties of dendrites for pattern discrimination and learning? What are the consequences of connecting such elements together in recurrent circuits? Do they still exhibit attractor dynamics? (See work of [http://lnc.usc.edu/publications.htm Bartlett Mel] to get started here, esp. papers from early 90's and early 2000's. )

	* '''NETtalk.''' Train a multi-layer perceptron to convert text to speech. You can get Sejnowski & Rosenberg's original paper and the data they used [http://cnl.salk.edu/Research/ParallelNetsPronounce/ here]. (You will need a DECtalk speech synthesizer to play the phonemes - you may be able to pick up a used one online.)		* '''NETtalk.''' Train a multi-layer perceptron to convert text to speech. You can get Sejnowski & Rosenberg's original paper and the data they used [http://cnl.salk.edu/Research/ParallelNetsPronounce/ here]. (You will need a DECtalk speech synthesizer to play the phonemes - you may be able to pick up a used one online.)
Line 24:		Line 24:

	* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)		* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)
			<!-- * '''Hardware implementation of associative memory.''' The analog Hopfield model has a direct physical implementation as an electrical circuit of resistors, capacitors, and op-amps. Try building a scaled-down version of this model in hardware. What issues arise in the implementation of this model? How long does it take to converge to a local minimum? -->
	* '''Hardware implementation of associative memory.''' The analog Hopfield model has a direct physical implementation as an electrical circuit of resistors, capacitors, and op-amps. Try building a scaled-down version of this model in hardware. What issues arise in the implementation of this model? How long does it take to converge to a local minimum?		<!-- * '''Analysis of Hopfield dynamics.''' The [http://redwood.berkeley.edu/vs265/hopfield84.pdf Hopfield dynamics] may be seen as performing a form of gradient descent on the energy function - i.e., the state of each unit follows a monotonically increasing function of the gradient rather than the gradient itself. The same is true of the [http://redwood.berkeley.edu/vs265/rozell-sparse-coding-nc08.pdf LCA] (locally competitive algorithm) used for sparse coding. Is the resulting trajectory more efficient for reaching an energy minimum than what you would get from doing steepest descent? -->

	* '''Analysis of Hopfield dynamics.''' The [http://redwood.berkeley.edu/vs265/hopfield84.pdf Hopfield dynamics] may be seen as performing a form of gradient descent on the energy function - i.e., the state of each unit follows a monotonically increasing function of the gradient rather than the gradient itself. The same is true of the [http://redwood.berkeley.edu/vs265/rozell-sparse-coding-nc08.pdf LCA] (locally competitive algorithm) used for sparse coding. Is the resulting trajectory more efficient for reaching an energy minimum than what you would get from doing steepest descent?

	* '''Bump circuits.''' Implement Kechen Zhang's [http://redwood.berkeley.edu/vs265/zhang96.pdf bump circuit model] discussed in class. How robust is the model to perturbations of the weights? How might such a circuit be made robust and self-correct for any imperfections in the weights?		* '''Bump circuits.''' Implement Kechen Zhang's [http://redwood.berkeley.edu/vs265/zhang96.pdf bump circuit model] discussed in class. How robust is the model to perturbations of the weights? How might such a circuit be made robust and self-correct for any imperfections in the weights?
Line 48:		Line 46:

	* '''Oscillations.''' Oscillations in neural activity are pervasive throughout the brain. What kinds of neural circuits are capable of eliciting oscillating behavior in spiking neurons? How could it be coordinated across large regions of cortex? (ask Fritz Sommer) What role might it play in the processing of information? [http://redwood.berkeley.edu/vs265/hopfield95.pdf John Hopfield] has suggested that spike timing relative to the phase of an ongoing oscillation could code information. See also [http://redwood.berkeley.edu/vs265/koepsell-sommer.pdf Koepsell et al. (2010)]. What factors would need to be considered in order to make this idea viable?		* '''Oscillations.''' Oscillations in neural activity are pervasive throughout the brain. What kinds of neural circuits are capable of eliciting oscillating behavior in spiking neurons? How could it be coordinated across large regions of cortex? (ask Fritz Sommer) What role might it play in the processing of information? [http://redwood.berkeley.edu/vs265/hopfield95.pdf John Hopfield] has suggested that spike timing relative to the phase of an ongoing oscillation could code information. See also [http://redwood.berkeley.edu/vs265/koepsell-sommer.pdf Koepsell et al. (2010)]. What factors would need to be considered in order to make this idea viable?
			<!--
	* '''Memristors.''' The memristor was originally proposed by Leon Chua at UC Berkeley back in the 1970's as a hypothetical '4th circuit element.' Recently, a group at [http://redwood.berkeley.edu/vs265/memristor-nature08.pdf HP labs] has discovered a memristive like element (see also [http://spectrum.ieee.org/semiconductors/processors/how-we-found-the-missing-memristor this article]). What is intriguing about the memristor is its synapse like properties ([http://redwood.berkeley.edu/vs265/memristor-stdp.pdf ref]), and according to Chua it even leads to a more straightforward model of action potential dynamics! Might memristive elements exist in neural circuits in the brain? Try simulating a memristive device and explore its relevance to plasticity or dynamics in neural systems.		* '''Memristors.''' The memristor was originally proposed by Leon Chua at UC Berkeley back in the 1970's as a hypothetical '4th circuit element.' Recently, a group at [http://redwood.berkeley.edu/vs265/memristor-nature08.pdf HP labs] has discovered a memristive like element (see also [http://spectrum.ieee.org/semiconductors/processors/how-we-found-the-missing-memristor this article]). What is intriguing about the memristor is its synapse like properties ([http://redwood.berkeley.edu/vs265/memristor-stdp.pdf ref]), and according to Chua it even leads to a more straightforward model of action potential dynamics! Might memristive elements exist in neural circuits in the brain? Try simulating a memristive device and explore its relevance to plasticity or dynamics in neural systems. -->

	* '''Learning in sensorimotor loops.''' A goal of research in both robotics and neuroscience is to understand the principles of adaptive behavior in embodied systems. However, currently there are few theories for guiding work in this area. One approach advanced by O'regan and colleagues is based on learning "sensorimotor contingencies" ([http://redwood.berkeley.edu/vs265/noe-oregan.pdf ref1], [http://redwood.berkeley.edu/vs265/philipona-oregan.pdf ref2]). Another by Ralf Der and colleagues is based on minimizing both predictive and 'postdictive' error ([http://redwood.berkeley.edu/vs265/taming_the_beast_martius2010.pdf ref]). Both propose simple algorithmic examples that you can implement in computer simulation, or you may wish to implement in a simple robotic platform.		* '''Learning in sensorimotor loops.''' A goal of research in both robotics and neuroscience is to understand the principles of adaptive behavior in embodied systems. However, currently there are few theories for guiding work in this area. One approach advanced by O'regan and colleagues is based on learning "sensorimotor contingencies" ([http://redwood.berkeley.edu/vs265/noe-oregan.pdf ref1], [http://redwood.berkeley.edu/vs265/philipona-oregan.pdf ref2]). Another by Ralf Der and colleagues is based on minimizing both predictive and 'postdictive' error ([http://redwood.berkeley.edu/vs265/taming_the_beast_martius2010.pdf ref]). Both propose simple algorithmic examples that you can implement in computer simulation, or you may wish to implement in a simple robotic platform.

Bruno at 01:13, 24 October 2014

2014-10-24T01:13:50Z

← Older revision		Revision as of 01:05, 24 October 2014
Line 1:		Line 1:
	'''Project presentations''' will take place on project presentation day~~, '''Tuesday December 11, 1-4pm'''~~. You may present your project either as an oral (15 min.) or poster presentation.		'''Project presentations''' will take place on project presentation day (TBD). You may present your project either as an oral (15 min.) or poster presentation.

	Think of the class project as an extended lab assignment. In most of the labs we just scratched the surface of various network models. This is a chance to explore these ideas in more depth. In addition to giving an oral or poster presentation on your project at the end of the term, you should turn in a short writeup that describes the problem you investigated, why it is interesting, your approach, results, and conclusions. ~~There is no length requirement but about 5-10 pages would be in the right ballpark.~~		Think of the class project as an extended lab assignment. In most of the labs we just scratched the surface of various network models. This is a chance to explore these ideas in more depth. In addition to giving an oral or poster presentation on your project at the end of the term, you should turn in a short writeup (5 pages) that describes the problem you investigated, why it is interesting, your approach, results, and conclusions.

	'''Project Suggestions'''		'''Project Suggestions'''

Bruno at 01:12, 24 October 2014

2014-10-24T01:12:12Z

← Older revision		Revision as of 01:04, 24 October 2014
Line 21:		Line 21:
	* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)		* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)

	* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: ~~{{math\|<VAR>τ</VAR>}}~~ dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.		* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.

	* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)		* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)
Line 27:		Line 27:
	* '''Hardware implementation of associative memory.''' The analog Hopfield model has a direct physical implementation as an electrical circuit of resistors, capacitors, and op-amps. Try building a scaled-down version of this model in hardware. What issues arise in the implementation of this model? How long does it take to converge to a local minimum?		* '''Hardware implementation of associative memory.''' The analog Hopfield model has a direct physical implementation as an electrical circuit of resistors, capacitors, and op-amps. Try building a scaled-down version of this model in hardware. What issues arise in the implementation of this model? How long does it take to converge to a local minimum?

	~~<!--~~ * '''Analysis of Hopfield dynamics.''' The [http://redwood.berkeley.edu/vs265/hopfield84.pdf Hopfield dynamics] may be seen as performing a form of gradient descent on the energy function - i.e., the state of each unit~~, <math>V_i</math>,~~ follows a monotonically increasing function of the gradient rather than the gradient itself. The same is true of the [http://redwood.berkeley.edu/vs265/rozell-sparse-coding-nc08.pdf LCA] (locally competitive algorithm) used for sparse coding. Is the resulting trajectory more efficient for reaching an energy minimum than what you would get from doing steepest descent? ~~-->~~		* '''Analysis of Hopfield dynamics.''' The [http://redwood.berkeley.edu/vs265/hopfield84.pdf Hopfield dynamics] may be seen as performing a form of gradient descent on the energy function - i.e., the state of each unit follows a monotonically increasing function of the gradient rather than the gradient itself. The same is true of the [http://redwood.berkeley.edu/vs265/rozell-sparse-coding-nc08.pdf LCA] (locally competitive algorithm) used for sparse coding. Is the resulting trajectory more efficient for reaching an energy minimum than what you would get from doing steepest descent?

	* '''Bump circuits.''' Implement Kechen Zhang's [http://redwood.berkeley.edu/vs265/zhang96.pdf bump circuit model] discussed in class. How robust is the model to perturbations of the weights? How might such a circuit be made robust and self-correct for any imperfections in the weights?		* '''Bump circuits.''' Implement Kechen Zhang's [http://redwood.berkeley.edu/vs265/zhang96.pdf bump circuit model] discussed in class. How robust is the model to perturbations of the weights? How might such a circuit be made robust and self-correct for any imperfections in the weights?

Bruno at 01:09, 24 October 2014

2014-10-24T01:09:48Z

← Older revision		Revision as of 01:01, 24 October 2014
Line 21:		Line 21:
	* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)		* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)

	* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: {{math\|<VAR>τ</VAR>}}dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.		* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: {{math\|<VAR>τ</VAR>}} dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.

	* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)		* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)

Bruno at 01:08, 24 October 2014

2014-10-24T01:08:06Z

← Older revision		Revision as of 01:00, 24 October 2014
Line 21:		Line 21:
	* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)		* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)

	* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.		* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: {{math\|<VAR>τ</VAR>}}dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.

	* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)		* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)

Bruno at 01:07, 24 October 2014

2014-10-24T01:07:03Z

← Older revision		Revision as of 00:59, 24 October 2014
Line 21:		Line 21:
	* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)		* '''The 'magic TV'.''' Suppose you woke up one day to find someone rewired your optic nerve (or you have been implanted with a prosthetic retina). The signals from retina to brain are intact, but the wires are all mixed up in the wrong place. Since neighboring pixels in natural images are correlated, it should be possible to learn a remapping that "descrambles" the image by exploiting these correlations. See if you can train a Kohonen-style network to learn the proper topographic mapping of an image based on the statistics of natural images. (Kohonen dubbed this problem 'the Magic TV'.)

	~~<!--~~ * '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: ~~<math>~~ y = W x~~</math>~~, or using just recurrent weights: ~~<math> \tau~~ dy/dt + y = x + M y~~</math>~~, or both: ~~<math>\tau~~ dy/dt + y = W x + M y~~</math>~~. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1. ~~-->~~		* '''Feedforward vs. recurrent weights.''' As we discussed in class, one can implement a given input-output mapping in a neural network using just feedforward weights: y = W x, or using just recurrent weights: dy/dt + y = x + M y, or both: dy/dt + y = W x + M y. Probably there is a trade-off here in terms of minimizing overall wiring length and settling time - i.e., feedforward networks are fast but require lots of synapses, while recurrent networks are slower but can implement more complex functions with local connections. Explore these tradeoffs for a particular problem - e.g., implementing an array of Gabor filters in a model of V1.

	* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)		* '''Sparse codes and associative memory.''' The advantages of storing and recalling patterns using an associative memory as opposed to conventional template matching are 1) parallel search, 2) distributed storage, and 3) denoising (recall of an uncorrupted pattern from partial or degraded input). However, associative memory models do not work well with natural data such as images or sound directly. Rather, they are best suited (have highest capacity) for sparse patterns (i.e., patterns with many zeros). Recent work (discussed in class) has shown how it is possible to convert natural images and sounds into a sparse format, and there is some evidence for this happening in the brain. See if you can link these ideas in order to store natural images or sounds in an associative memory. (You can read the work of [http://redwood.berkeley.edu/vs265/rehn-sommer-sparse-memory.pdf Rehn and Sommer] for one approach.)