StARformer: Transformer With State-Action-Reward Representations for Robot Learning.

Reinforcement Learning (RL) can be considered as a sequence modeling task, where an agent employs a sequence of past state-action-reward experiences to predict a sequence of future actions. In this work, we propose State-Action-Reward Transformer (StARformer), a Transformer architecture for robot learning with image inputs, which explicitly models short-term state-action-reward representations (StAR-representations), essentially introducing a Markovian-like inductive bias to improve long-term modeling. StARformer first extracts StAR-representations using self-attending patches of image states, action, and reward tokens within a short temporal window. These StAR-representations are combined with pure image state representations, extracted as convolutional features, to perform self-attention over the whole sequence. Our experimental results show that StARformer outperforms the state-of-the-art Transformer-based method on image-based Atari and DeepMind Control Suite benchmarks, under both offline-RL and imitation learning settings. We find that models can benefit from our combination of patch-wise and convolutional image embeddings. StARformer is also more compliant with longer sequences of inputs than the baseline method. Finally, we demonstrate how StARformer can be successfully applied to a real-world robot imitation learning setting via a human-following task.

Spherically Stratified Point Projection: Feature Image Generation for Object Classification Using 3D LiDAR Data.

Three-dimensional point clouds have been utilized and studied for the classification of objects at the environmental level. While most existing studies, such as those in the field of computer vision, have detected object type from the perspective of sensors, this study developed a specialized strategy for object classification using LiDAR data points on the surface of the object. We propose a method for generating a spherically stratified point projection (sP2) feature image that can be applied to existing image-classification networks by performing pointwise classification based on a 3D point cloud using only LiDAR sensors data. The sP2's main engine performs image generation through spherical stratification, evidence collection, and channel integration. Spherical stratification categorizes neighboring points into three layers according to distance ranges. Evidence collection calculates the occupancy probability based on Bayes' rule to project 3D points onto a two-dimensional surface corresponding to each stratified layer. Channel integration generates sP2 RGB images with three evidence values representing short, medium, and long distances. Finally, the sP2 images are used as a trainable source for classifying the points into predefined semantic labels. Experimental results indicated the effectiveness of the proposed sP2 in classifying feature images generated using the LeNet architecture.

The Korean Version of the University of California San Diego Performance-based Skills Assessment: Reliability and Validity.

OBJECTIVE: The study's aim was to develop and standardize a Korean version of the University of California San Diego Performance-based Skills Assessment (K-UPSA), which is used to evaluate the daily living function of patients with schizophrenia. METHODS: Study participants were 78 patients with schizophrenia and 27 demographically matched healthy controls. We evaluated the clinical states and cognitive functions to verify K-UPSA's reliability and validity. For clinical states, the Positive and Negative Syndrome Scale, Clinical Global Impression-Schizophrenia scale, and Social and Occupational Functioning Assessment Scale and Schizophrenia Quality of Life Scale-fourth revision were used. The Schizophrenia Cognition Rating Scale, Short-form of Korean-Wechsler Adult Intelligence Scale, and Wisconsin Card Sorting Test were used to assess cognitive function. RESULTS: The K-UPSA had statistically significant reliability and validity. The K-UPSA has high internal consistency (Cronbach's alpha, 0.837) and test-retest reliability (intra-class correlation coefficient, 0.381-0.792; p＜0.001). The K-UPSA had significant discriminant validity (p＜0.001). Significant correlations between the K-UPSA's scores and most of the scales and tests listed above demonstrated K-UPSA's concurrent validity (p＜0.001). CONCLUSION: The K-UPSA is useful to evaluate the daily living function in Korean patients with schizophrenia.

The F type mitochondrial genome of hard-shelled mussel: Mytilus coruscus (Mytiloida, Mytilidae).

We determined the complete mitochondrial genome of the female hard-shelled mussel Mytilus coruscus (Gould, 1869) (F type). The F type genome is composed of 16,642 bp in length including 12 protein coding genes (PCGs), 2 rRNA and 23 tRNA with the same gene content and order as the other Mytilus species which is characterized by the absence of the ATPase8 gene and addition of tRNA(Met) (AUA). The nucleotide composition of the genome shows that the percentage of A+T (63.2%) is higher than those of M. edulis complex species (M. edulis, M. galloprovincialis, M. trossulus) but lower than that of M. californianus. The F type mitochondrial genome of M. coruscus will provide useful information on the evolutionary history of the Mytilus species in the family Mitilidae.

The complete mitochondrial genome of the marine polychaete: Perinereis aibuhitensis (Phyllodocida, Nereididae).

We determined the complete mitochondrial genome sequence of the marine polychaete Perinereis aibuhitensis (Grube, 1878) (Phyllodocida, Nereididae), a dominant species in the mudflat of the West Pacific and Indian Ocean. The complete genome of P. aibuhitensis is 15,852 bp in size including 13 protein coding genes (PCGs), 2 rRNA and 22 tRNA with the same gene order and structure as those of other Nereididae species. The nucleotide composition is 29.5% A, 21.2% C, 14.1% G, 35.2% T, showing a high content of A+T with G being used least in the third codon position (6.7%). All PCGs use ATG as the start codon while for the stop codon COI and ND1 use incomplete codon of T. The mitogenome sequence of P. aibuhitensis is second to that of P. nuntia known in the genus Perinereis, which will provide useful information for understanding evolutionary history of the genus Perinereis within the family Nereididae.

Clinical effectiveness of the Kampo medicine kamishoyosan for adjunctive treatment of tardive dyskinesia in patients with schizophrenia: a 16-week open trial.

The purpose of the present study was to evaluate the clinical effectiveness of kamishoyosan for antipsychotic-induced tardive dyskinesia, and to investigate the relationship between tardive dyskinesia and serum brain-derived neurotrophic factor (BDNF) levels. Sixty-nine schizophrenia patients were enrolled; of these, 49 presented with tardive dyskinesia while the remaining 20 patients showed no tardive dyskinesia. The tardive dyskinesia group was treated for 16 weeks with kamishoyosan and assessed using the abnormal involuntary movement scale. The abnormal involuntary movement scale scores in the tardive dyskinesia group were evaluated at baseline and after 4, 8, and 16 weeks of treatment. The BDNF levels of all subjects were measured at baseline in order to compare differences in serum BDNF levels between the tardive dyskinesia group and the non-tardive dyskinesia group, and to correlate the severity of tardive dyskinesia and serum BDNF in the tardive dyskinesia group. A meaningful reduction in total abnormal involuntary movement scale scores was observed in the tardive dyskinesia group treated with kamishoyosan at 4, 8, and 16 weeks of treatment (P < 0.01). No significant differences in serum BDNF levels were detected between the tardive dyskinesia group and the non-tardive dyskinesia group at baseline. Furthermore, no significant correlation was seen between the severity of tardive dyskinesia and serum BDNF levels. The present study suggests that kamishoyosan might be a promising adjunctive treatment for antipsychotic-induced tardive dyskinesia.