Recurrent neural networks (RNNs) enable the production and processing of time-dependent signals such as those involved in movement and 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.
Code implementing the RFLO learning algorithm for the example shown in Figure 2 has been included as a source code file accompanying this manuscript.
- James M Murray
- James M Murray
- James M Murray
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
- Peter Latham, University College London, United Kingdom
© 2019, Murray
This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.
A hexanucleotide repeat expansion in C9ORF72 is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). A hallmark of ALS/FTD pathology is the presence of dipeptide repeat (DPR) proteins, produced from both sense GGGGCC (poly-GA, poly-GP, poly-GR) and antisense CCCCGG (poly-PR, poly-PG, poly-PA) transcripts. Translation of sense DPRs, such as poly-GA and poly-GR, depends on non-canonical (non-AUG) initiation codons. Here, we provide evidence for canonical AUG-dependent translation of two antisense DPRs, poly-PR and poly-PG. A single AUG is required for synthesis of poly-PR, one of the most toxic DPRs. Unexpectedly, we found redundancy between three AUG codons necessary for poly-PG translation. Further, the eukaryotic translation initiation factor 2D (EIF2D), which was previously implicated in sense DPR synthesis, is not required for AUG-dependent poly-PR or poly-PG translation, suggesting that distinct translation initiation factors control DPR synthesis from sense and antisense transcripts. Our findings on DPR synthesis from the C9ORF72 locus may be broadly applicable to many other nucleotide repeat expansion disorders.
The amyloid beta (Aβ) plaques found in Alzheimer’s disease (AD) patients’ brains contain collagens and are embedded extracellularly. Several collagens have been proposed to influence Aβ aggregate formation, yet their role in clearance is unknown. To investigate the potential role of collagens in forming and clearance of extracellular aggregates in vivo, we created a transgenic Caenorhabditis elegans strain that expresses and secretes human Aβ1-42. This secreted Aβ forms aggregates in two distinct places within the extracellular matrix. In a screen for extracellular human Aβ aggregation regulators, we identified different collagens to ameliorate or potentiate Aβ aggregation. We show that a disintegrin and metalloprotease a disintegrin and metalloprotease 2 (ADM-2), an ortholog of ADAM9, reduces the load of extracellular Aβ aggregates. ADM-2 is required and sufficient to remove the extracellular Aβ aggregates. Thus, we provide in vivo evidence of collagens essential for aggregate formation and metalloprotease participating in extracellular Aβ aggregate removal.