A tabular data generation framework guided by downstream tasks optimization.

Recently, generative models have been gradually emerging into the extended dataset field, showcasing their advantages. However, when it comes to generating tabular data, these models often fail to satisfy the constraints of numerical columns, which cannot generate high-quality datasets that accurately represent real-world data and are suitable for the intended downstream applications. Responding to the challenge, we propose a tabular data generation framework guided by downstream task optimization (TDGGD). It incorporates three indicators into each time step of diffusion generation, using gradient optimization to align the generated fake data. Unlike the traditional strategy of separating the downstream task model from the upstream data synthesis model, TDGGD ensures that the generated data has highly focused columns feasibility in upstream real tabular data. For downstream task, TDGGD strikes the utility of tabular data over solely pursuing statistical fidelity. Through extensive experiments conducted on real-world tables with explicit column constraints and tables without explicit column constraints, we have demonstrated that TDGGD ensures increasing data volume while enhancing prediction accuracy. To the best of our knowledge, this is the first instance of deploying downstream information into a diffusion model framework.

Direct and complete utilization of agricultural straw to fabricate all-biomass films with high-strength, high-haze and UV-shielding properties.

It is of vital significance to fabricate high-value-added materials from agricultural wastes by environmentally friendly and cost-effective processes. In this work, we propose an approach to directly and completely convert agricultural straw into multifunctional all-biomass films by introducing an entanglement network of additional cellulose to enhance the strength of the regenerated straw. First, natural wheat straw is dissolved in the ionic liquid 1-allyl-3-methylimidazolium chloride (AmimCl). Then, a small amount of cellulose with a high degree of polymerization (DP) is introduced to obtain straw/cellulose/AmimCl solutions, which are subsequently soaked in water for biomass regeneration, washed and dried to obtain straw/cellulose films. Dynamic shear rheological test confirms that after adding high-DP cellulose, an enhanced entanglement network forms in the solutions, which is essential to the processing and mechanical properties of materials. Extensional rheological test indicates that straw/cellulose/AmimCl solutions exhibit excellent spinnability and film-forming properties based on a significant increase in the capillary break-up time. Therefore, after regeneration in water, straw-based all-biomass films with high mechanical strength are obtained. When the content of additional wood pulp (WP, DP = 1300) with respect to total solids is 25 wt%, the obtained straw/WP all-biomass film reaches a tensile strength of 62 MPa. More interestingly, because there is no intentional chemical pretreatment and compositional isolation involved in this process, almost all of the components in straw, such as cellulose, lignin, hemicellulose and inorganic compounds, are retained in the final films. Thus, the resultant films have a superhigh haze of 97% while preventing 97% UVA (320-400 nm) and almost 100% UVB (280-320 nm). In sum, we demonstrate the complete and value-added utilization of low-grade bioresources by a facile, green and economical process to fabricate high-strength, high-haze and UV-shielding all-biomass films, which have great potential in low-cost, biodegradable and environmentally friendly packaging.