Search for: All records

Learning to perform a perceptual decision task is generally achieved through sessions of effortful practice with feedback. Here, we investigated how passive exposure to task-relevant stimuli, which is relatively effortless and does not require feedback, influences active learning. First, we trained mice in a sound-categorization task with various schedules combining passive exposure and active training. Mice that received passive exposure exhibited faster learning, regardless of whether this exposure occurred entirely before active training or was interleaved between active sessions. We next trained neural-network models with different architectures and learning rules to perform the task. Networks that use the statistical properties of stimuli to enhance separability of the data via unsupervised learning during passive exposure provided the best account of the behavioral observations. We further found that, during interleaved schedules, there is an increased alignment between weight updates from passive exposure and active training, such that a few interleaved sessions can be as effective as schedules with long periods of passive exposure before active training, consistent with our behavioral observations. These results provide key insights for the design of efficient training schedules that combine active learning and passive exposure in both natural and artificial systems. 

Recurrent neural networks (RNNs) enable the production and processing of time-dependent signals such as those involved in movement or working memory. Classic gradient-based algorithms for training RNNs have been available for decades, but are inconsistent with biological features of the brain, such as causality and locality. We derive an approximation to gradient-based learning that comports with these constraints by requiring synaptic weight updates to depend only on local information about pre- and postsynaptic activities, in addition to a random feedback projection of the RNN output error. In addition to providing mathematical arguments for the effectiveness of the new learning rule, we show through simulations that it can be used to train an RNN to perform a variety of tasks. Finally, to overcome the difficulty of training over very large numbers of timesteps, we propose an augmented circuit architecture that allows the RNN to concatenate short-duration patterns into longer sequences. 

Abstract The Eastern Pacific Warm Pool (EPWP) modulates global climate through its connection with tropical Pacific circulation, but sparse paleoceanographic data from this region limits our understanding of its role in past climate variability. We present a 144 kyr alkenone‐sea surface temperature (SST) reconstruction from core NH22P, located in the northern EPWP, that shows local warming occurred during periods of global cooling. Climate model simulations of the Last Glacial Maximum indicate that both ice sheet and greenhouse gas forcing slowed wind speeds over the EPWP, which attenuated glacial cooling of local SST via the wind‐evaporation‐SST feedback. Spectral analysis further suggests precessional pacing of the warming spikes. Vernal equinox insolation could explain this pacing as direct shortwave heating during boreal spring would have contributed to the early seasonal intensification of the EPWP. This work provides crucial constraints on tropical Pacific glacial climate variability and highlights the unique response of the EPWP to global climate forcings. 

The trace metal geochemistry of atmospheric dust and terrestrial surface particles were studied on the Qatar Peninsula from February 2014 to November 2015. We included samples of the mega dust-storm event on 01–02 April 2015. Atmospheric dust samples were collected using passive dust traps. Terrestrial surface deposits of recent dust accumulation and traffic particulate from roads were also sampled. All samples were total acid digested and analyzed for major and trace elements using ICP-OES analyzer. The concentration of thirteen elements (Ca, Mg, Ag, As, Cd, Cr, Cu, Mo, Ni, Se, Sn, Sr, Zn) were enriched in atmospheric dust samples, relative to upper continental crust (UCC). Calcium was especially enriched by up to 435% relative to UCC. About 33% of the total sample mass was CaCO3, reflecting the composition of surface rocks and soils in the source areas. Of the elements typically associated with anthropogenic activity, Ag, Ni, and Zn were most enriched relative to UCC, with enrichment factors (EF) of 182%, 233%, and 209%, respectively. Other metals, which normally reflect anthropogenic sources, including Pb and V, were not significantly enriched, with enrichment factors of 25% and 3%, respectively. Major elements (Al, Mn, Fe) were depleted (− 58%, − 35%, and − 5%, respectively) relative to UCC due to the large dilution effect of the enrichment of CaCO3. Back trajectories were determined at the date of sampling for each sample using the NOAA HYSPLIT model. These showed that the source of the dust particles was almost equally divided between northerly and southerly sources, except one sample, which appeared to originate from the west. More variability in particle source locations were observed during the winter months (October to March). Samples from the mega-dust storm were solubilized using an acetic acid-hydroxylamine hydrochloride leach procedure to obtain an upper estimate of the potential contribution of bioactive elements to surface seawater. The leach procedure solubilized a significant fraction of almost all elements. Ca was the element most affected (81% removed) because of the carbonate minerals present. Bioactive elements like Fe (25%) and P (58%) were also significantly solubilized. Because river input is so small to the Arabian Gulf, this solubilized fraction of dust is likely a major source of nutrients to surface seawater. Enrichment factors were also calculated with respect to the average composition of terrestrial surface deposits (TSD). Samples are not enriched significantly with respect to major components (EF < 2), with a depletion in Ca, K, Na in dust storm samples, reflecting a different origin. A significant enrichment of the same trace metals is evident in dust deposits and in traffic samples, but not in dust storms: Cu, Mo, Ni, Zn, possibly deriving from local atmospheric sources (traffic, industries). Samples with northern and southern origins were compared to see if the composition could be used to identify source. Only three elements were observed to be statistically different. Pb and Na were higher in samples from the south, while Cr was higher in those from the north. 

Abstract We explore the response of northeastern Pacific sea surface temperature (SST) to deglacial (16–7 ka) climate variability as recorded in‐based SST reconstructions spanning 65°N to 10°S. Included in the analysis is a new 23 kyr SST record from core NH8P from the northwest Mexican Margin. We isolate spatiotemporal patterns in regional SSTs with trend empirical orthogonal function (TEOF) analysis. The dominant TEOF mode reflects deglacial warming associated with rising. Tropical and subtropical SSTs correlated most strongly with this mode, suggesting that the thermodynamic response of the tropical eastern Pacific to greenhouse gas forcing was the dominant driver of regional SST change during deglaciation. The second TEOF mode reflects millennial‐scale variability and is most strongly expressed in subpolar SSTs. The synchronous timing between North Pacific and North Atlantic SST oscillations is evidence for the rapid transmission of millennial‐scale climate perturbations between the basins, likely through an atmospheric teleconnection. SSTs at NH8P have no correlation with either leading TEOF mode as there is minimal change in SST at this site after20 ka. A model simulation of the LGM indicates that glacial cooling was muted in much of the Eastern Pacific Warm Pool (EPWP), in which NH8P lies, due to reductions in latent heat flux. This suggests that the wind‐evaporation‐SST feedback was responsible for the attenuation of EPWP cooling. Overall, this study highlights the distinct latitudinal trends in the Pacific's response to deglaciation.