We examine how the saccade mechanism from biological vision can be used to make deep neural networks more efficient for classification and object detection problems. Our proposed approach is based on the ideas of attention-driven visual processing and saccades, miniature eye movements influenced by attention.
David Zhang
SRI Author
Real-Time Hyper-Dimensional Reconfiguration at the Edge using Hardware Accelerators
In this paper we present Hyper-Dimensional Reconfigurable Analytics at the Tactical Edge using low-SWaP embedded hardware that can perform real-time reconfiguration at the edge leveraging non-MAC deep neural nets (DNN) combined with hyperdimensional (HD) computing accelerators.
Hyper-Dimensional Analytics of Video Action at the Tactical Edge
We review HyDRATE, a low-SWaP reconfigurable neural network architecture developed under the DARPA AIE HyDDENN (Hyper-Dimensional Data Enabled Neural Network) program.
Bit Efficient Quantization for Deep Neural Networks
Quantization for deep neural networks have afforded models for edge devices that use less on-board memory and enable efficient low-power inference. In this paper, we present a comparison of model-parameter driven quantization approaches that can achieve as low as 3-bit precision without affecting accuracy. The post-training quantization approaches are data-free, and the resulting weight values are closely tied to the dataset distribution on which the model has converged to optimality. We show quantization results for a number of state-of-art deep neural networks (DNN) using large dataset like ImageNet. To better analyze quantization results, we describe the overall range and local sparsity of values afforded through various quantization schemes. We show the methods to lower bit-precision beyond quantization limits with object class clustering.
Fast, Full Chip Image Stitching of Nanoscale Integrated Circuits
The rapid progression of semiconductor technology has significantly impacted the ability to examine and analyze complex integrated circuits (ICs). Small device feature sizes, combined with large die sizes, add a heavy processing burden that severely limits our timely ability to perform defect validation and anti-tampering analysis at full scale. In this paper, we describe the algorithmic steps taken in the processing pipeline to quickly create a global image database of an entire advanced IC. We focused specifically on the image alignment and stitching algorithms necessary to support a combined field-of-view of a given layer of a die. We describe key algorithmic challenges such as contextual semantics that limits the robustness of the alignment algorithm. We also describe the use of database indexing to manage and traverse the enormous amounts of data.
Role for the Nicotinic Cholinergic System in Movement Disorders; Therapeutic Implications
A large body of evidence using experimental animal models shows that the nicotinic cholinergic system is involved in the control of movement under physiological conditions. This work raised the question whether dysregulation of this system may contribute to motor dysfunction and whether drugs targeting nicotinic acetylcholine receptors (nAChRs) may be of therapeutic benefit in movement disorders. Accumulating preclinical studies now show that drugs acting at nAChRs improve drug-induced dyskinesias. The general nAChR agonist nicotine, as well as several nAChR agonists (varenicline, ABT-089 and ABT-894), reduces l-dopa-induced abnormal involuntary movements or dyskinesias up to 60% in parkinsonian nonhuman primates and rodents. These dyskinesias are potentially debilitating abnormal involuntary movements that arise as a complication of l-dopa therapy for Parkinson’s disease. In addition, nicotine and varenicline decrease antipsychotic-induced abnormal involuntary movements in rodent models of tardive dyskinesia. Antipsychotic-induced dyskinesias frequently arise as a side effect of chronic drug treatment for schizophrenia, psychosis and other psychiatric disorders. Preclinical and clinical studies also show that the nAChR agonist varenicline improves balance and coordination in various ataxias. Lastly, nicotine has been reported to attenuate the dyskinetic symptoms of Tourette’s disorder. Several nAChR subtypes appear to be involved in these beneficial effects of nicotine and nAChR drugs including α4β2*, α6β2* and α7 nAChRs (the asterisk indicates the possible presence of other subunits in the receptor). Overall, the above findings, coupled with nicotine’s neuroprotective effects, suggest that nAChR drugs have potential for future drug development for movement disorders.
ABT-089 and ABT-894 REducation Divisionce Levodopa-Induced Dyskinesias in a Monkey Model of Parkinson’s Disease
Levodopa-induced dyskinesias (LIDs) are a serious complication of levodopa therapy for Parkinson’s disease for which there is little treatment. Accumulating evidence shows that nicotinic acetylcholine receptor (nAChR) drugs decrease LIDs in parkinsonian animals. Here, we examined the effect of two β2 nAChR agonists, ABT-089 and ABT-894, that previously were approved for phase 2 clinical trials for other indications. Two sets of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-lesioned monkeys were administered levodopa/carbidopa (10 mg/kg and 2.5 mg/kg, respectively) twice daily 5 days a week until they were stably dyskinetic. Each set had a vehicle-treated group, an nAChR agonist-treated group, and a nicotine-treated group as a positive control. Set A monkeys had previously received other nAChR drugs (nAChR drug-primed), whereas Set B monkeys were initially nAChR drug-naive. Both sets were administered the partial agonist ABT-089 (range, 0.01-1.0 mg/kg) orally 5 days a week twice daily 30 minutes before levodopa with each dose given for 1 to 5 weeks. ABT-089 decreased LIDs by 30% to 50% compared with vehicle-treated monkeys. Nicotine reduced LIDs by 70% in a parallel group. After 4 weeks of washout, the effect of the full agonist ABT-894 (range, 0.0001-0.10 mg/kg) was assessed on LIDs in Set A and Set B. ABT-894 reduced LIDs by 70%, similar to nicotine. Both drugs acted equally well at α4β2* and α6β2* nAChRs; however, ABT-089 was 30 to 60 times less potent than ABT-894. Tolerance did not develop for the time periods tested (range, 3-4 months). The nAChR drugs did not worsen parkinsonism or cognitive ability. Emesis, a common problem with nAChR drugs, was not observed. ABT-894 and ABT-089 appear to be good candidate nAChR drugs for the management of LIDs in Parkinson’s disease.
Nicotinic Receptor Agonists REducation Divisionce L-Dopa-Induced Dyskinesias in a Monkey Model of Parkinson’s Disease
Abstract Abnormal involuntary movements or dyskinesias are a serious complication of long-term l-DOPA treatment of Parkinson’s disease, for which there are few treatment options. Accumulating preclinical data show that nicotine decreases l-DOPA–induced dyskinesias (LIDs), suggesting that it may be a useful antidyskinetic therapy for Parkinson’s disease. Here, we investigated whether nicotinic acetylcholine receptor (nAChR) agonists reduced LIDs […]
Video Object Segmentation through Spatially Accurate and Temporally Dense Extraction of Primary Object Regions
In this paper, we propose a novel approach to extract primary object segments in videos in the ‘object proposal’ domain. The extracted primary object regions are then used to build object models for optimized video segmentation. The proposed approach has several contributions: First, a novel layered Directed Acyclic Graph (DAG) based frame-work is presented for detection and segmentation of the primary object in video. We exploit the fact that, in general, objects are spatially cohesive and characterized by locally smooth motion trajectories, to extract the primary object from the set of all available proposals based on motion, appearance and predicted-shape similarity across frames. Second, the DAG is initialized with an enhanced object proposal set where motion based proposal predictions (from adjacent frames) are used to expand the set of object proposals for a particular frame. Last, the paper presents a motion scoring function for selection of object proposal- that emphasizes high optical flow gradients at proposal boundaries to discriminate between moving objects and the background. The proposed approach is evaluated using several challenging benchmark videos and it outperforms both unsupervised and supervised state-of-the-art methods.
