[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-127886-en":3,"doc-seo-127886-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},127886,2336474459895,"Aria","https://ap-avatar.wpscdn.com/avatar/22000baeef7a5ed0655?x-image-process=image/resize,m_fixed,w_180,h_180&k=1786071322749376916",8,"Research & Report","Modular, Hierarchical Machine Learning for Sequential Goal Completion - Research Paper Summary","A robot performs sequential goal completion in a maze where objects appear in changing locations across trials. Monolithic neural network approaches require extensive retraining when goals or goal states change, whereas the proposed modular approach isolates optimizable components that can be reconfigured. The work combines cognitive map learners with hyperdimensional computing to encode CML node states into high-dimensional vectors for symbolic reasoning, enabling localized updates and near-optimal navigation without global retraining.","Modular, Hierarchical Machine Learning for Sequential Goal Completion  \nNathan R. McDonald*a  \naAir Force Research Laboratory, Information Directorate, 525 Brooks Road, Rome, NY 13441, USA  \nABSTRACT  \nGiven a maze populated with different objects, one may task a robot with a sequential goal completion task, e.g. 1) pickup a key then 2) unlock the door then 3) unlock the treasure chest. A typical machine learning (ML) solution would involve a monolithically trained artificial neural network (ANN) . However, if the sequence of goals or the goals themselves change, then the ANN must be significantly (or, at worst, completely) retrained. Instead of a monolithic ANN, a modular ML component would be 1) independently optimizable (task-agnostic) and 2) arbitrarily reconfigurable with other ML modules. This work describes a modular, hierarchical ML framework by integrating two emerging ML techniques: 1) cognitive map learners (CML) and 2) hyperdimensional computing (HDC) . A CML is a collection of three single layer ANNs (matrices) collaboratively trained to learn the topology of an abstract graph. Here, two CMLs were constructed, one describing locations on in 2D physical space and the other the relative distribution of objects found in this space. Each CML node states was encoded as a high-dimensional vector to utilize HDC, an ML algebra, for symbolic reasoning over these high-dimensional “symbol” vectors. In this way, each sub-goal above was described by algebraic equations ofCML node states. Multiple, independently trained CMLs were subsequently assembled together to navigate a maze to solve a sequential goal task. Critically, changes to these goals required only localized changes in the CMLHDC architecture, as opposed to a global ANN retraining scheme. This framework therefore enabled a more traditional engineering approach to ML, akin to digital logic design.  \nKeywords: hyperdimensional computing, vector symbolic architectures, cognitive map learners, artificial neural networks, neuroengineering, path planning, modular machine learning  \n1. INTRODUCTION  \nSince deep neural networks (DNN) are typically trained monolithically, or end-to-end, to solve a well-defined task, sequential goal completion tasks are difficult. With respect to maze puzzles, DNNs struggle with learning from minimal training data, knowledge transfer, generalization to novel environments, and generating human-interpretable models [1] . From a classical mathematics perspective, many planning tasks can be formulated as finding the shortest path on an abstract graph [2] . However, the standard Dijkstra and A* shortest path algorithms must compute the entire route before deciding the first step. If the goal location changes before the algorithms finishes, then the whole algorithm must be restarted from scratch. Such path planners are then not ideal for autonomous, expeditionary robotics, which must rapidly respond to dynamic environments.  \nAlternatively, a modular machine learning (ML) approach could segregate knowledge in modules, e.g. movement, spatial relationships, and positions. By encoding this information and decision space in a consistent information representation, multiple neural network modules can be independently prepared (learned or calculated) and integrated together into a larger assembly, in a manner similar to digital logic. This work demonstrated a modular, hierarchical ML framework implemented according to two emerging ML techniques: 1) cognitive map learners and 2) hyperdimensional computing.  \nCognitive map learners (CML) are a new approach to artificial neural networks (ANN) and are trained to learn the topology of an abstract graph [3] . The CML’s three separate yet collaboratively trained single-layer ANNs (matrices) each learn internal representations of a different aspect of the graph: 1) node states, 2) edge actions, and 3) edge action availability. As a result of this atypical segregation of information, the CML, though never explici","cbCaijJAXQmFQcHf","https://ap.wps.com/l/cbCaijJAXQmFQcHf","pdf",1040106,1,15,"English","en",105,"# Abstract\n# Introduction\n## Challenges for monolithic deep neural networks in sequential planning\n## Modular machine learning via segregated knowledge representations\n# Cognitive Map Learners (CML)\n## Three collaboratively trained single-layer networks\n## Near-optimal path computation via graph topology\n# Hyperdimensional Computing (HDC)\n## Vector symbolic architectures for module orchestration\n# Maze Path Planning Task\n## Object and grid CML integration with robot sensor inputs","[{\"question\":\"Why is sequential goal completion difficult for end-to-end deep neural networks?\",\"answer\":\"End-to-end models are trained monolithically for a fixed task, making them hard to adapt when environments or goal locations change. They also struggle with limited training data, knowledge transfer, generalization, and producing human-interpretable models.\"},{\"question\":\"How does the proposed modular framework avoid global retraining when goals change?\",\"answer\":\"It uses independently trained CML modules integrated with HDC. Changes to goals require only localized adjustments in the CML-HDC architecture rather than retraining a single monolithic ANN.\"},{\"question\":\"What roles do cognitive map learners and hyperdimensional computing play together?\",\"answer\":\"CML learns graph topology and supports iterative near-optimal path computation between a specified start and target state. HDC encodes CML node states into high-dimensional vectors and uses an algebraic similarity mechanism to orchestrate multiple CMLs for navigation.\"}]","Modular, Hierarchical Machine Learning for Sequential Goal Completion - Research Paper Summary | PDF",1785942607,38,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"modular-hierarchical-machine-learning-for-sequential-goal-completion-research-paper-summary","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/modular-hierarchical-machine-learning-for-sequential-goal-completion-research-paper-summary/127886/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-23","2026-08-05",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"Why is sequential goal completion difficult for end-to-end deep neural networks?","Question",{"text":76,"@type":77},"End-to-end models are trained monolithically for a fixed task, making them hard to adapt when environments or goal locations change. They also struggle with limited training data, knowledge transfer, generalization, and producing human-interpretable models.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How does the proposed modular framework avoid global retraining when goals change?",{"text":81,"@type":77},"It uses independently trained CML modules integrated with HDC. Changes to goals require only localized adjustments in the CML-HDC architecture rather than retraining a single monolithic ANN.",{"name":83,"@type":74,"acceptedAnswer":84},"What roles do cognitive map learners and hyperdimensional computing play together?",{"text":85,"@type":77},"CML learns graph topology and supports iterative near-optimal path computation between a specified start and target state. HDC encodes CML node states into high-dimensional vectors and uses an algebraic similarity mechanism to orchestrate multiple CMLs for navigation.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]