Luis Bettencourt

Understanding the factors driving innovation in energy technologies is of critical importance to mitigating climate change and addressing other energy-related global challenges. Low levels of innovation, measured in terms of energy patent filings, were noted in the 1980s and 90s as an issue of concern and were attributed to limited investment in public and private research and development (R&D). Here we build a comprehensive global database of energy patents covering the period 1970-2009, which is unique in its temporal and geographical scope. Analysis of the data reveals a recent, marked departure from historical trends. A sharp increase in rates of patenting has occurred over the last decade, particularly in renewable technologies, despite continued low levels of R&D funding. To solve the puzzle of fast innovation despite modest R&D increases, we develop a model that explains the nonlinear response observed in the empirical data of technological innovation to various types of investment. The model reveals a regular relationship between patents, R&D funding, and growing markets across technologies, and accurately predicts patenting rates at different stages of technological maturity and market development. We show quantitatively how growing markets have formed a vital complement to public R&D in driving innovative activity. These two forms of investment have each leveraged the effect of the other in driving patenting trends over long periods of time.

Hierarchical networks are known to achieve high classification accuracy on difficult machine-learning tasks. For many applications, a clear explanation of why the data was classified a certain way is just as important as the classification itself. However, the complexity of hierarchical networks makes them ill-suited for existing explanation methods. We propose a new method, contribution propagation, that gives per-instance explanations of a trained network's classifications. We give theoretical foundations for the proposed method, and evaluate its correctness empirically. Finally, we use the resulting explanations to reveal unexpected behavior of networks that achieve high accuracy on visual object-recognition tasks using well-known data sets.

Humans have a dual nature. We are subject to the same natural laws and forces as other species yet dominate global ecology and exhibit enormous variation in energy use, cultural diversity, and apparent social organization. We suggest scientists tackle these challenges with a macroecological approach-using comparative statistical techniques to identify deep patterns of variation in large datasets and to test for causal mechanisms. We show the power of a metabolic perspective for interpreting these patterns and suggesting possible underlying mechanisms, one that focuses on the exchange of energy and materials within and among human societies and with the biophysical environment. Examples on human foraging ecology, life history, space use, population structure, disease ecology, cultural and linguistic diversity patterns, and industrial and urban systems showcase the power and promise of this approach.

In the primate's primary visual cortex (V1), cells are classified in terms of two categories: simple cells and complex cells, given their response properties. While simple cells respond strongly to gating and bar stimuli at a certain phase and location, responses of complex cells are insensitive to small translation of stimulus within the receptive field [1]. Inspired by the response properties of simple and complex cells in the primary visual cortex, we propose a computational network to learn the receptive fields of these cells, and address the development of translation invariance from a temporal sequence of natural images. A generative model with sparseness constraints is devised to minimize the energy of prediction errors. Each simple cell is modulated by a higher layer of complex cells in a multiplicative fashion, where a slowness property and a trace-like rule are enforced on complex cells, as the result of a temporal coherence soft constraint. Furthermore, non-negativity constraints of the latent cell variables and weight matrices are imposed to fit the known neurophysiology. We present an online gradient descent algorithm to train our model from natural image sequences, in which a pre-training strategy is used to initialize the weights. The developed connection weights show that complex cell outputs are directly proportional to quadratic forms of simple cell responses. Each receptive field of simple cells develop a Gabor-like orientation filter, and each complex cell pools similar simple cell receptive fields - in retinotopic and feature space - producing the locally-invariant representation.

The concepts of sustainable development have experienced extraordinary success since their advent in the 1980s. They are now an integral part of the agenda of governments and corporations, and their goals have become central to the mission of research laboratories and universities worldwide. However, it remains unclear how far the field has progressed as a scientific discipline, especially given its ambitious agenda of integrating theory, applied science, and policy, making it relevant for development globally and generating a new interdisciplinary synthesis across fields. To address these questions, we assembled a corpus of scholarly publications in the field and analyzed its temporal evolution, geographic distribution, disciplinary composition, and collaboration structure. We show that sustainability science has been growing explosively since the late 1980s when foundational publications in the field increased its pull on new authors and intensified their interactions. The field has an unusual geographic footprint combining contributions and connecting through collaboration cities and nations at very different levels of development. Its decomposition into traditional disciplines reveals its emphasis on the management of human, social, and ecological systems seen primarily from an engineering and policy perspective. Finally, we show that the integration of these perspectives has created a new field only in recent years as judged by the emergence of a giant component of scientific collaboration. These developments demonstrate the existence of a growing scientific field of sustainability science as an unusual, inclusive and ubiquitous scientific practice and bode well for its continued impact and longevity.

Can lateral connectivity in the primary visual cortex account for the time dependence and intrinsic task difficulty of human contour detection? To answer this question, we created a synthetic image set that prevents sole reliance on either low-level visual features or high-level context for the detection of target objects. Rendered images consist of smoothly varying, globally aligned contour fragments (amoebas) distributed among groups of randomly rotated fragments (clutter). The time course and accuracy of amoeba detection by humans was measured using a two-alternative forced choice protocol with self-reported confidence and variable image presentation time (20-200 ms), followed by an image mask optimized so as to interrupt visual processing. Measured psychometric functions were well fit by sigmoidal functions with exponential time constants of 30-91 ms, depending on amoeba complexity. Key aspects of the psychophysical experiments were accounted for by a computational network model, in which simulated responses across retinotopic arrays of orientation-selective elements were modulated by cortical association fields, represented as multiplicative kernels computed from the differences in pairwise edge statistics between target and distractor images. Comparing the experimental and the computational results suggests that each iteration of the lateral interactions takes at least ms of cortical processing time. Our results provide evidence that cortical association fields between orientation selective elements in early visual areas can account for important temporal and task-dependent aspects of the psychometric curves characterizing human contour perception, with the remaining discrepancies postulated to arise from the influence of higher cortical areas. Current computer vision algorithms reproducing the feed-forward features of the primate visual pathway still fall far behind the capabilities of human subjects in detecting objects in cluttered backgrounds. Here we investigate the possibility that recurrent lateral interactions, long hypothesized to form cortical association fields, can account for the dependence of object detection accuracy on shape complexity and image exposure time. Cortical association fields are thought to aid object detection by reinforcing global image features that cannot easily be detected by single neurons in feed-forward models. Our implementation uses the spatial arrangement, relative orientation, and continuity of putative contour elements to compute the lateral contextual support. We designed synthetic images that allowed us to control object shape and background clutter while eliminating unintentional cues to the presence of an otherwise hidden target. In contrast, real objects can vary uncontrollably in shape, are camouflaged to different degrees by background clutter, and are often associated with non-shape cues, making results using natural image sets difficult to interpret. Our computational model of cortical association fields matches many aspects of the time course and object detection accuracy of human subjects on statistically identical synthetic image sets. This implies that lateral interactions may selectively reinforce smooth object global boundaries.

Conventional sparse coding learns optimal dictionaries of feature bases to approximate input signals; however, it is not favorable to classify the inputs. Recent research has focused on building discriminative sparse coding models to facilitate the classification tasks. In this paper, we develop a new discriminative sparse coding model via bidirectional flows. Sensory inputs (from bottom-up) and discriminative signals (supervised from top-down) are propagated through a hierarchical network to form sparse representations at each level. The ℓ 0 -constrained sparse coding model allows highly efficient online learning and does not require iterative steps to reach a fixed point of the sparse representation. The introduction of discriminative top-down information flows helps to group reconstructive features belonging to the same class and thus to benefit the classification tasks. Experiments are conducted on multiple data sets including natural images, hand-written digits and 3-D objects with favorable results. Compared with unsupervised sparse coding via only bottom-up directions, the two-way discriminative approach improves the recognition performance significantly.

Salient features instantly attract visual attention to their location and are crucial for object recognition. Experiments in ultra-fast visual perception have shown that object recognition can be surprisingly accurate given only ∼20 ms of observation. Such short times exclude neural dynamics of top-down feedback and require fast mechanisms of low-level feature detection. We derive a neural model of the primary visual cortex with physiologically parameterized horizontal connections that reinforce salient features, and apply it to detect salient contours on ultra-fast time scales. Model performance qualitatively matches experimental results for human perception of contours, suggesting rapid neural mechanisms involving feedforward horizontal connections can be used to distinguish low-level objects.

With urban population increasing dramatically worldwide, cities are playing an increasingly critical role in human societies and the sustainability of the planet. An obstacle to effective policy is the lack of meaningful urban metrics based on a quantitative understanding of cities. Typically, linear per capita indicators are used to characterize and rank cities. However, these implicitly ignore the fundamental role of nonlinear agglomeration integral to the life history of cities. As such, per capita indicators conflate general nonlinear effects, common to all cities, with local dynamics, specific to each city, failing to provide direct measures of the impact of local events and policy. Agglomeration nonlinearities are explicitly manifested by the superlinear power law scaling of most urban socioeconomic indicators with population size, all with similar exponents ( 1.15). As a result larger cities are disproportionally the centers of innovation, wealth and crime, all to approximately the same degree. We use these general urban laws to develop new urban metrics that disentangle dynamics at different scales and provide true measures of local urban performance. New rankings of cities and a novel and simpler perspective on urban systems emerge. We find that local urban dynamics display long-term memory, so cities under or outperforming their size expectation maintain such (dis)advantage for decades. Spatiotemporal correlation analyses reveal a novel functional taxonomy of U.S. metropolitan areas that is generally not organized geographically but based instead on common local economic models, innovation strategies and patterns of crime.