2019 International Joint Conference on Neural Networks (IJCNN), 2019
Many temporal processing tasks face great challenges. The handcrafted features and structure used... more Many temporal processing tasks face great challenges. The handcrafted features and structure used in these temporal processing methods result in the brittle systems. To avoid handcrafted designs, we analyze and propose the EmergentContext Emergent-Input (ECEI) framework in this work. The Developmental Network (DN-1) can be considered as the first ECEI framework that used emergent motor vectors as temporal context (state) for top-down attention. This method suggests that the developmental root of temporal states is at the (open) motor end and always expressible by the motor end, not originally internal and hidden. Indeed, it has enabled the control of a Turing machine to incrementally emerge inside the network. However, DN-1 used a handcrafted partition of motor neurons as concept regions, which does not allow context to automatically emerge. The same is true for hidden (internal) areas other than the motor area. We analyze why this limitation is fundamentally damaging. To overcome this limitation, Developmental Network 2 (DN-2) is proposed with some new biology inspired mechanisms. DN2 is an ECEI framework which can directly use any naturally emerged motor patterns, allowing a freedom of thought that has been largely overlooked. The phoneme recognition experiments are used to investigate DN-2’s performance with two different settings about the motor area: handcrafted motor concept zones and emergent motor area. Based on the experimental results, we compared and analyzed such two mechanisms. This work is a step toward our goal of empowering DN-2 to not only free thoughts, but such free-thoughts are also necessary for handling increasingly complicated temporal processing tasks.
Approaches to computational intelligence, including neural networks, fuzzy systems and evolutiona... more Approaches to computational intelligence, including neural networks, fuzzy systems and evolutionary computation, are converging to a common frontier --autonomous mental development (AMD). This article explains what AMD is and why computational intelligence can fully expand its power at this frontier. As an example, this paper discusses a theory, an architecture, and some experimental results of AMD. Juyang Weng Michigan State University
The universal Turing Machine (TM) is a model for Von Neumann computers — general-purpose computer... more The universal Turing Machine (TM) is a model for Von Neumann computers — general-purpose computers. A human brain, linked with its biological body, can inside-skull-autonomously learn a universal TM so that he acts as a general-purpose computer and writes a computer program for any practical purposes. It is unknown whether a robot can accomplish the same. This theoretical work shows how the Developmental Network (DN), linked with its robot body, can accomplish this. Unlike a traditional TM, the TM learned by DN is a super TM — Grounded, Emergent, Natural, Incremental, Skulled, Attentive, Motivated, and ive (GENISAMA). A DN is free of any central controller (e.g., Master Map, convolution, or error back-propagation). Its learning from a teacher TM is one transition observation at a time, immediate, and error-free until all its neurons have been initialized by early observed teacher transitions. From that point on, the DN is no longer error-free but is always optimal at every time inst...
2015 International Joint Conference on Neural Networks (IJCNN), 2015
Typical navigation systems do not use object recognition as part of their autonomous driving syst... more Typical navigation systems do not use object recognition as part of their autonomous driving systems. People often use hand-crafted features (e.g. lanes, traffic lights, intersections) based on programmers' knowledge about the environment. Those agents are usually brittle during real-world tests. However, landmarks, as a type of object, need to be recognized for an autonomous navigation system to generalize its learned training data to other unfamiliar environments. In this work we utilize the Developmental Network (DN), which has been tested extensively with object recognition tasks, for a mobile agent and train it to self-navigate in controlled indoor environment. The proposed system uses Lobe Component Analysis (LCA) to learn features from both stereo cameras and desired navigation actions. Neurons attend to different areas of the input image. They compete for firing according to the goodness of matching result. This enables the agent to attend to image local patches as landmarks without explicitly defining objects. Our analysis shows that attention can be corrected by direct supervision and by indirect reinforcement provided by the teacher. We anticipate our work in this paper to be a starting point of research efforts that shift expensive range-scanner-based methods to inexpensive camera-based methods that, although using richer information, face challenges of object appearance variations.
ABSTRACT An effective framework for general object recognition and localization from complex back... more ABSTRACT An effective framework for general object recognition and localization from complex backgrounds had not been found till the brain-inspired Where-What Network (WWN) series by Weng and coworkers. This paper reports two advances along this line. One is the automatic adaptation of the receptive field of each neuron to disregard input dimensions that arise from backgrounds but without a handcrafted object model, since the initial hexagonal receptive field does not fit well the contour of the automatically assigned object view. The other is the hierarchical parallelization technique and its implementation on the GPU-based accelerator using the CUDA parallel language. The experimental results showed that automatic adaptation of the receptive fields led to improvements in the recognition rate. The hierarchical parallelization technique has achieved a speedup of 16 times compared to the C program. This speed-up was employed on the Haibao Robot displayed at the World Expo, Shanghai 2010.
The displacement field computed by the method discussed in the previous chapter provides point co... more The displacement field computed by the method discussed in the previous chapter provides point correspondences for motion and structure analysis. This chapter deals with estimating motion and structure of the scene from point correspondences between two perspective views. First, an algorithm is presented that computes a closed-form solution for motion and structure parameters. The algorithm exploits redundancy in the data to obtain more reliable estimates in the presence of noise. Then, an approach is introduced to estimating the errors in the computed solution. Specifically, standard deviation of the error is estimated in terms of the variance of the errors in the coordinates of the image points. The estimated errors indicate the reliability of the solution, as well as any degeneracy or near degeneracy that causes the failure of the motion estimation algorithm. The presented approach to error estimation is applicable to a wide variety of problems that involve least-squares optimization or pseudo-inverse. Finally, the relationships between errors and other parameters are analyzed.
Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)
The capability of learning new skills is very important for an artificial agent to scale up. In t... more The capability of learning new skills is very important for an artificial agent to scale up. In this paper, we propose a developmental cognitive learning architecture which enables an artificial agent to develop complex behaviors (chained actions) after acquisition of simple ones. The mechanism that makes this possible is chained secondary conditioning. The major challenge of this work is that
Using long stereo image sequences, a recursive-batch approach is used to fuse multiple stereo vie... more Using long stereo image sequences, a recursive-batch approach is used to fuse multiple stereo views. The objective of this recursive-batch approach is to achieve higher performance without suffering from excessive computational cost. The information about the object structure is preserved from one image frame to the next, and it is refined recursively. The structure is represented in a camera-centered and a world-centered coordinate system. The former is useful for continuous navigation and the latter for visual map generation. In the presence of noise, the expected accuracy is provided together with the estimated solutions. With this recursive-batch method, the estimated motion trajectory closely tracks actual motions and it has eliminated the long divergence time period, which typically covers about 20 to 30 image frames, with a conventional Kalman filtering method.
2019 International Joint Conference on Neural Networks (IJCNN), 2019
Many temporal processing tasks face great challenges. The handcrafted features and structure used... more Many temporal processing tasks face great challenges. The handcrafted features and structure used in these temporal processing methods result in the brittle systems. To avoid handcrafted designs, we analyze and propose the EmergentContext Emergent-Input (ECEI) framework in this work. The Developmental Network (DN-1) can be considered as the first ECEI framework that used emergent motor vectors as temporal context (state) for top-down attention. This method suggests that the developmental root of temporal states is at the (open) motor end and always expressible by the motor end, not originally internal and hidden. Indeed, it has enabled the control of a Turing machine to incrementally emerge inside the network. However, DN-1 used a handcrafted partition of motor neurons as concept regions, which does not allow context to automatically emerge. The same is true for hidden (internal) areas other than the motor area. We analyze why this limitation is fundamentally damaging. To overcome this limitation, Developmental Network 2 (DN-2) is proposed with some new biology inspired mechanisms. DN2 is an ECEI framework which can directly use any naturally emerged motor patterns, allowing a freedom of thought that has been largely overlooked. The phoneme recognition experiments are used to investigate DN-2’s performance with two different settings about the motor area: handcrafted motor concept zones and emergent motor area. Based on the experimental results, we compared and analyzed such two mechanisms. This work is a step toward our goal of empowering DN-2 to not only free thoughts, but such free-thoughts are also necessary for handling increasingly complicated temporal processing tasks.
Approaches to computational intelligence, including neural networks, fuzzy systems and evolutiona... more Approaches to computational intelligence, including neural networks, fuzzy systems and evolutionary computation, are converging to a common frontier --autonomous mental development (AMD). This article explains what AMD is and why computational intelligence can fully expand its power at this frontier. As an example, this paper discusses a theory, an architecture, and some experimental results of AMD. Juyang Weng Michigan State University
The universal Turing Machine (TM) is a model for Von Neumann computers — general-purpose computer... more The universal Turing Machine (TM) is a model for Von Neumann computers — general-purpose computers. A human brain, linked with its biological body, can inside-skull-autonomously learn a universal TM so that he acts as a general-purpose computer and writes a computer program for any practical purposes. It is unknown whether a robot can accomplish the same. This theoretical work shows how the Developmental Network (DN), linked with its robot body, can accomplish this. Unlike a traditional TM, the TM learned by DN is a super TM — Grounded, Emergent, Natural, Incremental, Skulled, Attentive, Motivated, and ive (GENISAMA). A DN is free of any central controller (e.g., Master Map, convolution, or error back-propagation). Its learning from a teacher TM is one transition observation at a time, immediate, and error-free until all its neurons have been initialized by early observed teacher transitions. From that point on, the DN is no longer error-free but is always optimal at every time inst...
2015 International Joint Conference on Neural Networks (IJCNN), 2015
Typical navigation systems do not use object recognition as part of their autonomous driving syst... more Typical navigation systems do not use object recognition as part of their autonomous driving systems. People often use hand-crafted features (e.g. lanes, traffic lights, intersections) based on programmers' knowledge about the environment. Those agents are usually brittle during real-world tests. However, landmarks, as a type of object, need to be recognized for an autonomous navigation system to generalize its learned training data to other unfamiliar environments. In this work we utilize the Developmental Network (DN), which has been tested extensively with object recognition tasks, for a mobile agent and train it to self-navigate in controlled indoor environment. The proposed system uses Lobe Component Analysis (LCA) to learn features from both stereo cameras and desired navigation actions. Neurons attend to different areas of the input image. They compete for firing according to the goodness of matching result. This enables the agent to attend to image local patches as landmarks without explicitly defining objects. Our analysis shows that attention can be corrected by direct supervision and by indirect reinforcement provided by the teacher. We anticipate our work in this paper to be a starting point of research efforts that shift expensive range-scanner-based methods to inexpensive camera-based methods that, although using richer information, face challenges of object appearance variations.
ABSTRACT An effective framework for general object recognition and localization from complex back... more ABSTRACT An effective framework for general object recognition and localization from complex backgrounds had not been found till the brain-inspired Where-What Network (WWN) series by Weng and coworkers. This paper reports two advances along this line. One is the automatic adaptation of the receptive field of each neuron to disregard input dimensions that arise from backgrounds but without a handcrafted object model, since the initial hexagonal receptive field does not fit well the contour of the automatically assigned object view. The other is the hierarchical parallelization technique and its implementation on the GPU-based accelerator using the CUDA parallel language. The experimental results showed that automatic adaptation of the receptive fields led to improvements in the recognition rate. The hierarchical parallelization technique has achieved a speedup of 16 times compared to the C program. This speed-up was employed on the Haibao Robot displayed at the World Expo, Shanghai 2010.
The displacement field computed by the method discussed in the previous chapter provides point co... more The displacement field computed by the method discussed in the previous chapter provides point correspondences for motion and structure analysis. This chapter deals with estimating motion and structure of the scene from point correspondences between two perspective views. First, an algorithm is presented that computes a closed-form solution for motion and structure parameters. The algorithm exploits redundancy in the data to obtain more reliable estimates in the presence of noise. Then, an approach is introduced to estimating the errors in the computed solution. Specifically, standard deviation of the error is estimated in terms of the variance of the errors in the coordinates of the image points. The estimated errors indicate the reliability of the solution, as well as any degeneracy or near degeneracy that causes the failure of the motion estimation algorithm. The presented approach to error estimation is applicable to a wide variety of problems that involve least-squares optimization or pseudo-inverse. Finally, the relationships between errors and other parameters are analyzed.
Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)
The capability of learning new skills is very important for an artificial agent to scale up. In t... more The capability of learning new skills is very important for an artificial agent to scale up. In this paper, we propose a developmental cognitive learning architecture which enables an artificial agent to develop complex behaviors (chained actions) after acquisition of simple ones. The mechanism that makes this possible is chained secondary conditioning. The major challenge of this work is that
Using long stereo image sequences, a recursive-batch approach is used to fuse multiple stereo vie... more Using long stereo image sequences, a recursive-batch approach is used to fuse multiple stereo views. The objective of this recursive-batch approach is to achieve higher performance without suffering from excessive computational cost. The information about the object structure is preserved from one image frame to the next, and it is refined recursively. The structure is represented in a camera-centered and a world-centered coordinate system. The former is useful for continuous navigation and the latter for visual map generation. In the presence of noise, the expected accuracy is provided together with the estimated solutions. With this recursive-batch method, the estimated motion trajectory closely tracks actual motions and it has eliminated the long divergence time period, which typically covers about 20 to 30 image frames, with a conventional Kalman filtering method.
Uploads