Israel Nelken - Academia.edu (original) (raw)

Papers by Israel Nelken

Probabilistic modeling of temporal phenomena is of central importance in a variety of fields rang... more Probabilistic modeling of temporal phenomena is of central importance in a variety of fields ranging from neuroscience to economics to speech recognition.

The goal of most learning processes is to bring a machine into a set of "correct&quo... more The goal of most learning processes is to bring a machine into a set of "correct" states. In practice, however, it may be difficult to show that the process enters this target set. We present a condition that ensures that the process visits the target set infinitely often almost surely. This condition is easy to verify and is true for many well-known learning rules. To demonstrate the utility of this method, we apply it to four types of learning processes: the perceptron, learning rules governed by continuous energy functions, the Kohonen rule, and the committee machine.

The explosive nature of a focal cortical seizure suggests the operation of strong positive feedba... more The explosive nature of a focal cortical seizure suggests the operation of strong positive feedback in the neuronal network. It was previously proposed that in the hippocampus this may be provided by the regenerative accumulation of potassium in the interstitium. This hypothesis was severely criticized in the past. However, it gained new impetus with the recent discovery that focal seizures can arise in mammalian hippocampal slices perfused with low calcium solutions despite the block of chemical synaptic transmission. Here, we examine the relationship of interstitial potassium concentration to the electrogenesis of these low-calcium seizures. Both the experimental data and the behavior of a simplified mathematical model describing neuronal discharge in low calcium, support the contention that interstitial potassium accumulation may play an important role in the buildup of hippocampal seizures.

The minimal change in a stimulus property that is detectable by neurons has been often quantified... more The minimal change in a stimulus property that is detectable by neurons has been often quantified using the receiver operating characteristic (ROC) curve, but recent studies introduced the use of the related Fisher information (FI). Whereas ROC analysis and FI quantify the information available for discriminating between two stimuli, global aspects of the information carried by a neuron are quantified by the mutual information (MI) between stimuli and responses. FI and MI have been shown to be related to each other when FI is large. Here the responses of neurons recorded in the inferior colliculus of anesthetized guinea pigs in response to ensembles of sounds differing in their interaural time differences (ITDs) or binaural correlation (BC) were analyzed. Although the FI is not uniformly large, there are strong relationships between MI and FI. Information-theoretic measures are used to demonstrate the importance of the non-Poisson statistics of these responses. These neurons may reflect the maximization of the MI between stimuli and responses under constraints on the coded stimulus range and the range of firing rates. Remarkably, whereas the maximization of MI, in conjunction with the non-Poisson statistics of the spike trains, is enough to create neurons whose ITD discrimination capabilities are close to the behavioral limits, the same rule does not achieve single-neuron BC discrimination that is as close to behavioral performance.

Auditory neurons that exhibit stimulus-specific adaptation (SSA) decrease their response to commo... more Auditory neurons that exhibit stimulus-specific adaptation (SSA) decrease their response to common tones while retaining responsiveness to rare ones. We recorded single-unit responses from the inferior colliculus (IC) where SSA is known to occur and we explored for the first time SSA in the cochlear nucleus (CN) of rats. We assessed an important functional outcome of SSA, the extent to which frequency discriminability depends on sensory context. For this purpose, pure tones were presented in an oddball sequence as standard (high probability of occurrence) or deviant (low probability of occurrence) stimuli. To study frequency discriminability under different probability contexts, we varied the probability of occurrence and the frequency separation between tones. The neuronal sensitivity was estimated in terms of spike-count probability using signal detection theory. We reproduced the finding that many neurons in the IC exhibited SSA, but we did not observe significant SSA in our CN sample. We concluded that strong SSA is not a ubiquitous phenomenon in the CN. As predicted, frequency discriminability was enhanced in IC when stimuli were presented in an oddball context, and this enhancement was correlated with the degree of SSA shown by the neurons. In contrast, frequency discrimination by CN neurons was independent of stimulus context. Our results demonstrated that SSA is not widespread along the entire auditory pathway, and suggest that SSA increases frequency discriminability of single neurons beyond that expected from their tuning curves.

This paper presents a new way of modeling the activity of single neurons in stochastic settings. ... more This paper presents a new way of modeling the activity of single neurons in stochastic settings. It incorporates in a natural way many physiological mechanisms not usually found in stochastic models, such as spatial integration, non-linear membrane characteristics and non-linear interactions between excitation and inhibition. The model is based on the fact that most of the neuronal inputs have a finite lifetime. Thus, the stochastic input can be modeled as a simple finite markov chain, and the membrane potential becomes a function of the state of this chain. Firing occurs at states whose membrane potential is above threshold. The main mathematical results of the model are: (i) the input-output firing rate curve is convex at low firing rates and is saturated at high firing rates, and (ii) at low firing rates, firing usually occurs when there is synchronous convergence of many excitatory events.

Fishbach, Alon, Israel Nelken, and Yehezkel Yeshurun. Auditory edge detection: a neural model for... more Fishbach, Alon, Israel Nelken, and Yehezkel Yeshurun. Auditory edge detection: a neural model for physiological and psychoacoustical responses to amplitude transients. J Neurophysiol 85: 2001. Primary segmentation of visual scenes is based on spatiotemporal edges that are presumably detected by neurons throughout the visual system. In contrast, the way in which the auditory system decomposes complex auditory scenes is substantially less clear. There is diverse physiological and psychophysical evidence for the sensitivity of the auditory system to amplitude transients, which can be considered as a partial analogue to visual spatiotemporal edges. However, there is currently no theoretical framework in which these phenomena can be associated or related to the perceptual task of auditory source segregation. We propose a neural model for an auditory temporal edge detector, whose underlying principles are similar to classical visual edge detector models. Our main result is that this model reproduces published physiological responses to amplitude transients collected at multiple levels of the auditory pathways using a variety of experimental procedures. Moreover, the model successfully predicts physiological responses to a new set of amplitude transients, collected in cat primary auditory cortex and medial geniculate body. Additionally, the model reproduces several published psychoacoustical responses to amplitude transients as well as the psychoacoustical data for amplitude edge detection reported here for the first time. These results support the hypothesis that the response of auditory neurons to amplitude transients is the correlate of psychoacoustical edge detection.

Every time we listen--to speech, to music, to footsteps approaching or retreating--our auditory p... more Every time we listen--to speech, to music, to footsteps approaching or retreating--our auditory perception is the result of a long chain of diverse and intricate processes that unfold within the source of the sound itself, in the air, in our ears, and, most of all, in our brains. ...

Animal models of MMN may serve both to further our understanding of neural processing beyond pure... more Animal models of MMN may serve both to further our understanding of neural processing beyond pure sensory coding and for unraveling the neural and pharmacological processes involved in the generation of MMN. We start this review by discussing the methodological issues that are especially important when pursuing a single-neuron correlate of MMN. Correlates of MMN have been studied in mice,

Nature Neuroscience, 2009

Studies of auditory cortex are often driven by the assumption, derived from our better understand... more Studies of auditory cortex are often driven by the assumption, derived from our better understanding of visual cortex, that basic physical properties of sounds are represented there before being used by higher-level areas for determining sound-source identity and location. However, we only have a limited appreciation of what the cortex adds to the extensive subcortical processing of auditory information, which can account for many perceptual abilities. This is partly because of the approaches that have dominated the study of auditory cortical processing to date, and future progress will unquestionably profit from the adoption of methods that have provided valuable insights into the neural basis of visual perception. At the same time, we propose that there are unique operating principles employed by the auditory cortex that relate largely to the simultaneous and sequential processing of previously derived features and that therefore need to be studied and understood in their own right.

Sound processing by the auditory system is understood in unprecedented details, even compared wit... more Sound processing by the auditory system is understood in unprecedented details, even compared with sensory coding in the visual system. Nevertheless, we do not understand yet the way in which some of the simplest perceptual properties of sounds are coded in neuronal activity. This poses serious difficulties for linking neuronal responses in the auditory system and music processing, since music operates on abstract representations of sounds. Paradoxically, although perceptual representations of sounds most probably occur high in auditory system or even beyond it, neuronal responses are strongly affected by the temporal organization of sound streams even in subcortical stations. Thus, to the extent that music is organized sound, it is the organization, rather than the sound, which is represented first in the auditory brain.

Probabilistic modeling of temporal phenomena is of central importance in a variety of fields rang... more Probabilistic modeling of temporal phenomena is of central importance in a variety of fields ranging from neuroscience to economics to speech recognition.

The goal of most learning processes is to bring a machine into a set of "correct&quo... more The goal of most learning processes is to bring a machine into a set of "correct" states. In practice, however, it may be difficult to show that the process enters this target set. We present a condition that ensures that the process visits the target set infinitely often almost surely. This condition is easy to verify and is true for many well-known learning rules. To demonstrate the utility of this method, we apply it to four types of learning processes: the perceptron, learning rules governed by continuous energy functions, the Kohonen rule, and the committee machine.

The explosive nature of a focal cortical seizure suggests the operation of strong positive feedba... more The explosive nature of a focal cortical seizure suggests the operation of strong positive feedback in the neuronal network. It was previously proposed that in the hippocampus this may be provided by the regenerative accumulation of potassium in the interstitium. This hypothesis was severely criticized in the past. However, it gained new impetus with the recent discovery that focal seizures can arise in mammalian hippocampal slices perfused with low calcium solutions despite the block of chemical synaptic transmission. Here, we examine the relationship of interstitial potassium concentration to the electrogenesis of these low-calcium seizures. Both the experimental data and the behavior of a simplified mathematical model describing neuronal discharge in low calcium, support the contention that interstitial potassium accumulation may play an important role in the buildup of hippocampal seizures.

The minimal change in a stimulus property that is detectable by neurons has been often quantified... more The minimal change in a stimulus property that is detectable by neurons has been often quantified using the receiver operating characteristic (ROC) curve, but recent studies introduced the use of the related Fisher information (FI). Whereas ROC analysis and FI quantify the information available for discriminating between two stimuli, global aspects of the information carried by a neuron are quantified by the mutual information (MI) between stimuli and responses. FI and MI have been shown to be related to each other when FI is large. Here the responses of neurons recorded in the inferior colliculus of anesthetized guinea pigs in response to ensembles of sounds differing in their interaural time differences (ITDs) or binaural correlation (BC) were analyzed. Although the FI is not uniformly large, there are strong relationships between MI and FI. Information-theoretic measures are used to demonstrate the importance of the non-Poisson statistics of these responses. These neurons may reflect the maximization of the MI between stimuli and responses under constraints on the coded stimulus range and the range of firing rates. Remarkably, whereas the maximization of MI, in conjunction with the non-Poisson statistics of the spike trains, is enough to create neurons whose ITD discrimination capabilities are close to the behavioral limits, the same rule does not achieve single-neuron BC discrimination that is as close to behavioral performance.

Auditory neurons that exhibit stimulus-specific adaptation (SSA) decrease their response to commo... more Auditory neurons that exhibit stimulus-specific adaptation (SSA) decrease their response to common tones while retaining responsiveness to rare ones. We recorded single-unit responses from the inferior colliculus (IC) where SSA is known to occur and we explored for the first time SSA in the cochlear nucleus (CN) of rats. We assessed an important functional outcome of SSA, the extent to which frequency discriminability depends on sensory context. For this purpose, pure tones were presented in an oddball sequence as standard (high probability of occurrence) or deviant (low probability of occurrence) stimuli. To study frequency discriminability under different probability contexts, we varied the probability of occurrence and the frequency separation between tones. The neuronal sensitivity was estimated in terms of spike-count probability using signal detection theory. We reproduced the finding that many neurons in the IC exhibited SSA, but we did not observe significant SSA in our CN sample. We concluded that strong SSA is not a ubiquitous phenomenon in the CN. As predicted, frequency discriminability was enhanced in IC when stimuli were presented in an oddball context, and this enhancement was correlated with the degree of SSA shown by the neurons. In contrast, frequency discrimination by CN neurons was independent of stimulus context. Our results demonstrated that SSA is not widespread along the entire auditory pathway, and suggest that SSA increases frequency discriminability of single neurons beyond that expected from their tuning curves.

This paper presents a new way of modeling the activity of single neurons in stochastic settings. ... more This paper presents a new way of modeling the activity of single neurons in stochastic settings. It incorporates in a natural way many physiological mechanisms not usually found in stochastic models, such as spatial integration, non-linear membrane characteristics and non-linear interactions between excitation and inhibition. The model is based on the fact that most of the neuronal inputs have a finite lifetime. Thus, the stochastic input can be modeled as a simple finite markov chain, and the membrane potential becomes a function of the state of this chain. Firing occurs at states whose membrane potential is above threshold. The main mathematical results of the model are: (i) the input-output firing rate curve is convex at low firing rates and is saturated at high firing rates, and (ii) at low firing rates, firing usually occurs when there is synchronous convergence of many excitatory events.

Fishbach, Alon, Israel Nelken, and Yehezkel Yeshurun. Auditory edge detection: a neural model for... more Fishbach, Alon, Israel Nelken, and Yehezkel Yeshurun. Auditory edge detection: a neural model for physiological and psychoacoustical responses to amplitude transients. J Neurophysiol 85: 2001. Primary segmentation of visual scenes is based on spatiotemporal edges that are presumably detected by neurons throughout the visual system. In contrast, the way in which the auditory system decomposes complex auditory scenes is substantially less clear. There is diverse physiological and psychophysical evidence for the sensitivity of the auditory system to amplitude transients, which can be considered as a partial analogue to visual spatiotemporal edges. However, there is currently no theoretical framework in which these phenomena can be associated or related to the perceptual task of auditory source segregation. We propose a neural model for an auditory temporal edge detector, whose underlying principles are similar to classical visual edge detector models. Our main result is that this model reproduces published physiological responses to amplitude transients collected at multiple levels of the auditory pathways using a variety of experimental procedures. Moreover, the model successfully predicts physiological responses to a new set of amplitude transients, collected in cat primary auditory cortex and medial geniculate body. Additionally, the model reproduces several published psychoacoustical responses to amplitude transients as well as the psychoacoustical data for amplitude edge detection reported here for the first time. These results support the hypothesis that the response of auditory neurons to amplitude transients is the correlate of psychoacoustical edge detection.

Every time we listen--to speech, to music, to footsteps approaching or retreating--our auditory p... more Every time we listen--to speech, to music, to footsteps approaching or retreating--our auditory perception is the result of a long chain of diverse and intricate processes that unfold within the source of the sound itself, in the air, in our ears, and, most of all, in our brains. ...

Animal models of MMN may serve both to further our understanding of neural processing beyond pure... more Animal models of MMN may serve both to further our understanding of neural processing beyond pure sensory coding and for unraveling the neural and pharmacological processes involved in the generation of MMN. We start this review by discussing the methodological issues that are especially important when pursuing a single-neuron correlate of MMN. Correlates of MMN have been studied in mice,

Nature Neuroscience, 2009

Studies of auditory cortex are often driven by the assumption, derived from our better understand... more Studies of auditory cortex are often driven by the assumption, derived from our better understanding of visual cortex, that basic physical properties of sounds are represented there before being used by higher-level areas for determining sound-source identity and location. However, we only have a limited appreciation of what the cortex adds to the extensive subcortical processing of auditory information, which can account for many perceptual abilities. This is partly because of the approaches that have dominated the study of auditory cortical processing to date, and future progress will unquestionably profit from the adoption of methods that have provided valuable insights into the neural basis of visual perception. At the same time, we propose that there are unique operating principles employed by the auditory cortex that relate largely to the simultaneous and sequential processing of previously derived features and that therefore need to be studied and understood in their own right.

Sound processing by the auditory system is understood in unprecedented details, even compared wit... more Sound processing by the auditory system is understood in unprecedented details, even compared with sensory coding in the visual system. Nevertheless, we do not understand yet the way in which some of the simplest perceptual properties of sounds are coded in neuronal activity. This poses serious difficulties for linking neuronal responses in the auditory system and music processing, since music operates on abstract representations of sounds. Paradoxically, although perceptual representations of sounds most probably occur high in auditory system or even beyond it, neuronal responses are strongly affected by the temporal organization of sound streams even in subcortical stations. Thus, to the extent that music is organized sound, it is the organization, rather than the sound, which is represented first in the auditory brain.