Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (6 results)

  • Thesis
  • Peer Reviewed
UC Berkeley Electronic Theses and Dissertations (2019)

Animals function in a 3D world in which survival depends on robust, well-controlled actions. Historically, researchers in Artificial Intelligence (AI) and neuroscience have explored sensory and motor systems independently. There is a growing body of literature in AI and neuroscience to suggest that they actually work in tandem. While there has been a great deal of work on vision and audition as sensory modalities in these fields, one could argue that a more fundamental modality in biology is haptics, or the sense of touch. In this thesis, we will look at building computational models that integrate tactile sensing with other sensory modalities to perform manipulation-like tasks in robots and discrimination tasks in mice. We will also explore the problem of inference through the lens of Markov Chain Monte Carlo methods (MCMC). We elaborate on the ideas discussed in this thesis in the introduction presented in Chapter 1.

A challenging problem one often faces when applying probabilistic mathematical models to the study of sensory-motor systems and other problems involving learning of inference is sampling. Hamiltonian Markov Chain Monte Carlo (HMC) algorithms can efficiently draw representative samples from complex probabilistic models. Most MCMC methods rely on detailed balance to ensure that we can sample from the correct distribution. This constraint can be relaxed in discrete state spaces such as those employed by HMC type methods. In Chapter 2, we study HMC methods without detailed balance to explore faster convergence. Markov jump processes are stochastic processes on discrete state space but continuous in time. In Chapter 3, we use Markov Jump Processes to simulate waiting times along with generalized detailed balance. This waiting time ,we show, helps generate samples faster. Most MCMC methods are plagued by slow simulation times on discrete computing systems. In Chapter 4, we explore HMC in analog circuits where the problem of generating samples from a distribution is mapped to the problem of sampling charge in a capacitor.

The second half of this dissertation focuses on the role of haptics in perception and action. Manipulation is a fundamental problem for artificial and biological agents. High dimensional actuators (say, fingers, trunks,etc) are really hard to control. In Chapter 5, we present an approach to learn to actuate dexterous manipulators to grasp objects in simulation. Haptics as a sensory modality is critical to many manipulation tasks. Employing haptics in high dimensional dextrous actuators is challenging. In Chapter 6, we explore how intrinsic curiosity and haptics can be used to learn exploration strategies for discrimination of objects with dextrous hands. A key component to make tactile sensing a possibility is the availability of cheap, efficient, scalable hardware. Chapter 7 presents results for tactile servoing using a physical gelsight sensor. Traditional neuroscience texts delineate sensory and motor systems as two independent systems yet recent results suggest that this may not be entirely complete. That is, there is evidence to suggest that the representations in the cortex is more distributed than is accepted. Finally in Chapter 8, we explore building a computational model of spiking neural data collected from both the barrel and motor cortices during free and active whisking. These works help towards understanding sensorimotor representations in the context of haptics and high dimensional controls. We conclude with a discussion on future directions in Chapter 9.

    Cover page: Inference Algorithms and Sensorimotor Representations in Brains and Machines
    • Article
    • Peer Reviewed
    UC Berkeley Previously Published Works (2018)
    We extract pixel-level masks of extreme weather patterns using variants of Tiramisu and DeepLabv3+ neural networks. We describe improvements to the software frameworks, input pipeline, and the network training algorithms necessary to efficiently scale deep learning on the Piz Daint and Summit systems. The Tiramisu network scales to 5300 P100 GPUs with a sustained throughput of 21.0 PF/s and parallel efficiency of 79.0%. DeepLabv3+ scales up to 27360 V100 GPUs with a sustained throughput of 325.8 PF/s and a parallel efficiency of 90.7% in single precision. By taking advantage of the FP16 Tensor Cores, a half-precision version of the DeepLabv3+ network achieves a peak and sustained throughput of 1.13 EF/s and 999.0 PF/s respectively.
      Cover page: Exascale Deep Learning for Climate Analytics
      • Article
      • Peer Reviewed
      LBL Publications (2018)
      For the past year, the HEP.TrkX project has been investigating machine learning solutions to LHC particle track reconstruction problems. A variety of models were studied that drew inspiration from computer vision applications and operated on an image-like representation of tracking detector data. While these approaches have shown some promise, image-based methods face challenges in scaling up to realistic HL-LHC data due to high dimensionality and sparsity. In contrast, models that can operate on the spacepoint representation of track measurements ("hits") can exploit the structure of the data to solve tasks efficiently. In this paper we will show two sets of new deep learning models for reconstructing tracks using space-point data arranged as sequences or connected graphs. In the first set of models, Recurrent Neural Networks (RNNs) are used to extrapolate, build, and evaluate track candidates akin to Kalman Filter algorithms. Such models can express their own uncertainty when trained with an appropriate likelihood loss function. The second set of models use Graph Neural Networks (GNNs) for the tasks of hit classification and segment classification. These models read a graph of connected hits and compute features on the nodes and edges. They adaptively learn which hit connections are important and which are spurious. The models are scaleable with simple architecture and relatively few parameters. Results for all models will be presented on ACTS generic detector simulated data.
        Cover page: Novel deep learning methods for track reconstruction
        • Article
        • Peer Reviewed
        LBL Publications (2017)
        Particle track reconstruction in dense environments such as the detectors of the High Luminosity Large Hadron Collider (HL-LHC) is a challenging pattern recognition problem. Traditional tracking algorithms such as the combinatorial Kalman Filter have been used with great success in LHC experiments for years. However, these state-of-the-art techniques are inherently sequential and scale poorly with the expected increases in detector occupancy in the HL-LHC conditions. The HEP.TrkX project is a pilot project with the aim to identify and develop cross-experiment solutions based on machine learning algorithms for track reconstruction. Machine learning algorithms bring a lot of potential to this problem thanks to their capability to model complex non-linear data dependencies, to learn effective representations of high-dimensional data through training, and to parallelize easily on high-throughput architectures such as GPUs. This contribution will describe our initial explorations into this relatively unexplored idea space. We will discuss the use of recurrent (LSTM) and convolutional neural networks to find and fit tracks in toy detector data.
          Cover page: The HEP.TrkX Project: deep neural networks for HL-LHC online and offline tracking
          • Article
          • Peer Reviewed
          UC Berkeley Previously Published Works (2021)
          Identifying, detecting, and localizing extreme weather events is a crucial first step in understanding how they may vary under different climate change scenarios. Pattern recognition tasks such as classification, object detection, and segmentation (i.e., pixel-level classification) have remained challenging problems in the weather and climate sciences. While there exist many empirical heuristics for detecting extreme events, the disparities between the output of these different methods even for a single event are large and often difficult to reconcile. Given the success of deep learning (DL) in tackling similar problems in computer vision, we advocate a DL-based approach. DL, however, works best in the context of supervised learning - when labeled datasets are readily available. Reliable labeled training data for extreme weather and climate events is scarce. We create "ClimateNet" - an open, community-sourced human-expert-labeled curated dataset that captures tropical cyclones (TCs) and atmospheric rivers (ARs) in highresolution climate model output from a simulation of a recent historical period.We use the curated ClimateNet dataset to train a state-of-the-art DL model for pixel-level identification - i.e., segmentation - of TCs and ARs. We then apply the trained DL model to historical and climate change scenarios simulated by the Community Atmospheric Model (CAM5.1) and show that the DL model accurately segments the data into TCs, ARs, or "the background" at a pixel level. Further, we show how the segmentation results can be used to conduct spatially and temporally precise analytics by quantifying distributions of extreme precipitation conditioned on event types (TC or AR) at regional scales. The key contribution of this work is that it paves the way for DL-based automated, high-fidelity, and highly precise analytics of climate data using a curated expert-labeled dataset - ClimateNet. ClimateNet and the DL-based segmentation method provide several unique capabilities: (i) they can be used to calculate a variety of TC and AR statistics at a fine-grained level; (ii) they can be applied to different climate scenarios and different datasets without tuning as they do not rely on threshold conditions; and (iii) the proposed DL method is suitable for rapidly analyzing large amounts of climate model output. While our study has been conducted for two important extreme weather patterns (TCs and ARs) in simulation datasets, we believe that this methodology can be applied to a much broader class of patterns and applied to observational and reanalysis data products via transfer learning.
            Cover page: ClimateNet: an expert-labeled open dataset and deep learning architecture for enabling high-precision analyses of extreme weather
            • Article
            • Peer Reviewed
            LBL Publications (2020)
            Identifying, detecting and localizing extreme weather events is a crucial first step in understanding how they may vary under different climate change scenarios. Pattern recognition tasks such as classification, object detection and segmentation have remained challenging problems in the weather and climate sciences. While there exist many empirical heuristics for detecting extreme events, the disparities between the output of these different methods even for a single event are large and often difficult to reconcile. Given the success of Deep Learning (DL) in tackling similar problems in computer vision, we advocate a DL-based approach. DL, however, works best in the context of supervised learning; when labeled datasets are readily available. Reliable, labeled training data for extreme weather and climate events is scarce. We create ClimateNet – an open, community-sourced human expert-labeled curated dataset – that captures tropical cyclones (TCs) and atmospheric rivers (ARs) in high-resolution climate model output from a simulation of a recent historical period. We use the curated ClimateNet dataset to train a state-of-the-art DL model for pixel-level identification, i.e. segmentation, of TCs and ARs. We then apply the trained DL model to historical and climate change scenarios simulated by the Community Atmospheric Model (CAM5.1) and show that the DL model accurately segments the data into TCs, ARs or the background at a pixel level. Further, we show how the segmentation results can be used to conduct spatially and temporally precise analytics by quantifying distributions of extreme precipitation conditioned on event types (TC or AR) at regional scales. The key contribution of this work is that it paves the way for DL-based automated, hi-fidelity and highly precise analytics of climate data using a curated expert-labelled dataset – ClimateNet. ClimateNet and the DL-based segmentation method provide several unique capabilities: (i) they can be used to calculate a variety of TC and AR statistics at a fine-grained level; (ii) they can be applied to different climate scenarios and different datasets without tuning as they do not rely on threshold conditions; and (iii) the proposed DL method is suitable for rapidly analyzing large amounts of climate model output. While our study has been conducted for two important extreme weather patterns (TCs and ARs) in simulation datasets, we believe that this methodology can be applied to a much broader class of patterns, and applied to observational and reanalysis data products via transfer learning.
              Creative Commons 'BY' version 4.0 license
              Top