The in vivo firing patterns of ventral midbrain dopamine neurons are controlled by afferent and intrinsic activity to generate sensory cue and prediction error signals that are essential for reward-based learning. Given the absence of in vivo intracellular recordings during the last three decades, the subthreshold membrane potential events that cause changes in dopamine neuron firing patterns remain unknown. To address this, we established in vivo whole-cell recordings and obtained over 100 spontaneously active, immunocytochemically-defined midbrain dopamine neurons in isoflurane-anaesthetized adult mice. We identified a repertoire of subthreshold membrane potential signatures associated with distinct in vivo firing patterns. Dopamine neuron activity in vivo deviated from single-spike pacemaking by phasic increases in firing rate via two qualitatively distinct biophysical mechanisms: 1) a prolonged hyperpolarization preceding rebound bursts, accompanied by a hyperpolarizing shift in action potential threshold; and 2) a transient depolarization leading to high-frequency plateau bursts, associated with a depolarizing shift in action potential threshold. Our findings define a mechanistic framework for the biophysical implementation of dopamine neuron firing patterns in the intact brain.

Various intrusive and nonintrusive appliance load monitoring and classification systems have been studied; however, most of them designed so far provide group-level energy usage feedback. We present the first phase of a system with the potential to attribute energy-related events to an individual occupant of a space and provide occupant-specific energy usage feedback in an uninstrumented space (e.g., home or office). This initial phase focuses on collecting the electromagnetic field (EMF) radiated by several common appliances to determine a unique signature for each appliance. It also implements a machine learning algorithm to classify appliances from an incoming EMF data file. The proposed approach has been prototyped with hardware realization. The results obtained on tested appliances indicate the EMF sensor's ability and potential to develop a system for providing occupant-specific energy feedback.

The firing pattern of ventral midbrain dopamine neurons is controlled by afferent and intrinsic activity to generate prediction error signals that are essential for reward-based learning. Given the absence of intracellular in vivo recordings in the last three decades, the subthreshold membrane potential events that cause changes in dopamine neuron firing patterns remain unknown. By establishing stable in vivo whole-cell recordings of >100 spontaneously active midbrain dopamine neurons in anaesthetized mice, we identified the repertoire of subthreshold membrane potential signatures associated with distinct in vivo firing patterns. We demonstrate that in vivo activity of dopamine neurons deviates from a single spike pacemaker pattern by eliciting transient increases in firing rate generated by at least two diametrically opposing biophysical mechanisms: a transient depolarization resulting in high frequency plateau bursts associated with a reactive depolarizing shift in action potential threshold, and a prolonged hyperpolarization preceding slower rebound bursts characterized by a predictive hyperpolarizing shift in action potential threshold. Our findings therefore illustrate a framework for the biophysical implementation of prediction error coding in dopamine neurons by tuning action potential threshold dynamics.

Mobile phone metadata is increasingly used for humanitarian purposes in developing countries as traditional data is scarce. Basic demographic information is however often absent from mobile phone datasets, limiting the operational impact of the datasets. For these reasons , there has been a growing interest in predicting demographic information from mobile phone metadata. Previous work focused on creating increasingly advanced features to be modeled with standard machine learning algorithms. We here instead model the raw mobile phone meta-data directly using deep learning, exploiting the temporal nature of the patterns in the data. From high-level assumptions we design a data representation and convolutional network architecture for modeling patterns within a week. We then examine three strategies for aggregating patterns across weeks and show that our method reaches state-of-the-art accuracy on both age and gender prediction using only the temporal modality in mobile metadata. We finally validate our method on low activity users and evaluate the modeling assumptions.

No abstract

No abstract

In the field of songbird neuroscience, researchers have used playback of aversive noise bursts to drive changes in song behavior for specific syllables within a bird's song. Typically, a short (~5-10 msec) slice of the syllable is selected for targeting and the average spectrum of the slice is used as a template. Sounds that are sufficiently close to the template are considered a match. If other syllables have portions that are spectrally similar to the target, false positive errors will weaken the operant contingency. We present a gradient descent method for template optimization that increases the separation in distance between target and distractors slices, greatly improving targeting accuracy. Applied to songs from five adult Bengalese finches, the fractional reduction in errors for sub-syllabic slices was 51.54±22.92%. At the level of song syllables, we use an error metric that controls for the vastly greater number of distractors vs. target syllables. Setting 5% average error (misses + false positives) as a minimal performance criterion, the number of targetable syllables increased from 3 to 16 out of 61 syllables. At 10% error, targetable syllables increased from 11 to 26. By using simple and robust linear discriminant methods, the algorithm reaches near asymptotic performance when using 10 songs as training data, and the error increases by <2.3% on average when using only a single song for training. Targeting is temporally precise, with average jitter of 3.33 msec for the 16 accurately targeted syllables. Because the algorithm is concerned only with the problem of template selection, it can be used as a simple and robust front end for existing hardware and software implementations for triggered feedback.

AbstractIn the field of songbird neuroscience, researchers have used playback of aversive noise bursts to drive changes in song behavior for specific syllables within a bird’s song. Typically, a short (~5-10 msec) slice of the syllable is selected for targeting and the average spectrum of the slice is used as a template. Sounds that are sufficiently close to the template are considered a match. If other syllables have portions that are spectrally similar to the target, false positive errors will weaken the operant contingency. We present a gradient descent method for template optimization that increases the separation in distance between target and distractors slices, greatly improving targeting accuracy. Applied to songs from five adult Bengalese finches, the fractional reduction in errors for sub-syllabic slices was 51.54±22.92%. At the level of song syllables, we use an error metric that controls for the vastly greater number of distractors vs. target syllables. Setting 5% average error (misses + false positives) as a minimal performance criterion, the number of targetable syllables increased from 3 to 16 out of 61 syllables. At 10% error, targetable syllables increased from 11 to 26. By using simple and robust linear discriminant methods, the algorithm reaches near asymptotic performance when using 10 songs as training data, and the error increases by <2.3% on average when using only a single song for training. Targeting is temporally precise, with average jitter of 3.33 msec for the 16 accurately targeted syllables. Because the algorithm is concerned only with the problem of template selection, it can be used as a simple and robust front end for existing hardware and software implementations for triggered feedback.