Federated learning (FL) is a machine learning setting where many clients (e.g. mobile devices or whole organizations) collaboratively train a model under the orchestration of a central server (e.g. service provider), while keeping the training data decentralized. FL embodies the principles of focused data collection and minimization, and can mitigate many of the systemic privacy risks and costs resulting from traditional, centralized machine learning and data science approaches. Motivated by the explosive growth in FL research, this paper discusses recent advances and presents an extensive collection of open problems and challenges.

Shape memory polymers (SMPs), as a class of programmable stimuliresponsive shape changing polymers, are attracting increasing attention from the standpoint of both fundamental research and technological innovations. Following a brief introduction of the conventional shape memory effect (SME), progress in new shape memory enabling mechanisms and triggering methods, variations of in shape memory forms (shape memory surfaces, hydrogels, and microparticles), new shape memory behavior (multi-SME and two-way-SME), and novel fabrication methods are reviewed. Progress in thermomechanical modeling of SMPs is also presented. Abbreviations:SCPs shape changing polymers; LCEs liquid crystalline elastomers; SMP shape memory polymer; SME shape memory effect; 2W two-way; 1W oneway; T trans transition temperature; T g glass transition temperature; T m melting temperature; T cl liquid crystal cleaning temperature; T d deformation temperature; T f shape fixing temperature; T c crystallization temperature; R f shape stability ratio; R r shape recovery ratio; T sw switching temperature; ε max maximum recoverable strain; σ max maximum recovery stress; SMC shape memory cycle; ε load strain under load; ε fixed strain; T r recovery Page 2 of 106 A c c e p t e d M a n u s c r i p t 2 temperature; ε rec recovered strain; V r strain recovery rate; T σmax temperature corresponding to σ max ; T sw,app apparent switching temperature; PCL poly(ε-caprolactone); SMPU shape memory polyurethane; EMU elemental memory unit; TME temperature memory effect; PU polyurethane; T i liquid-crystal isotropic transition temperature; T v vitrification temperature; CIE crystallization induced elongation; MIC melting induced contraction

as dynamers by Lehn, [ 25,26 ] are stimuli-responsive polymers, most notably exhibiting macroscopic responses to changes in pH. [ 27,28 ] Several imine-containing polymers have been demonstrated, including pH-responsive hydrogels [ 20 ] and a working organic light-emitting diode (OLED). [ 23 ] However, the potential of polyimines as malleable, mechanically resilient polymeric materials, as well as their processability, have remained largely unexplored. We envision that imine-linked polymers can take malleability in covalent network polymers to the next level of simplicity, affordability and practicality. Herein, we present the fi rst catalyst-free malleable polyimine which fundamentally behaves like a classic thermoset at ambient conditions yet can be reprocessed by application of either heat or water. This means that green, room temperature processing conditions are accessible for this important class of functional polymers.A crosslinked polyimine network was prepared from commercially available monomers: terephthaldehyde, diethylene triamine, and triethylene tetramine ( Figure 1 a). A polyimine fi lm was obtained by simply mixing the three above components in a 3:0.9:1.4 stoichiometry in the absence of any catalyst in a mixture of organic solvents (1:1:8, v/v/v, CH 2 Cl 2 /EtOAc/EtOH), then allowing the volatiles to evaporate slowly. Alternatively, the polymer can be obtained as a powder by using ethyl acetate as the only solvent. The polymerization reaction was confi rmed by infrared spectroscopy, which revealed that aldehyde end groups were consumed while imine linkages were formed ( Figure S2, Supporting Information). The resulting translucent polymer is hard and glassy at room temperature ( T g is 56 °C) ( Figure S1, Supporting Information) and has a modulus of near 1 GPa with stress at break of 40 MPa ( Figure S3, Supporting Information).The time and temperature dependent relaxation modulus of the polyimine fi lm was tested to characterize the heat-induced malleability. Figure 1 b depicts the results of a series of relaxation tests over a wide range of temperatures (50-127.5 °C) on a double logarithmic plot. Specifi cally, at 80 °C, the bond exchange reaction is initiated and the normalized relaxation modulus is decreased from 1 to 0.11 within 30 min, indicating an 89% release of the internal stress within the thermoset polymer. By shifting each relaxation curve horizontally with respect to a reference temperature at 80 °C, a master relaxation curve was constructed (Figure 1 c), which indicates the stress relaxation of the polyimine follows the classic time-temperature superposition (TTSP) behavior. The plot of time-temperature shift factors as a function of temperature (Figure 1 d) shows that the polyimine's stress-relaxation behavior exhibits Arrheniuslike temperature dependence. Using the extrapolation, we calculated that while it takes 30 min for the stress to be relaxed by ca. 90% at 80 °C, the same process would take ca. 480 days at room temperature. The polyimine is thus the fi rst reported

No abstract